|Review of “The impact of atmospheric dynamics on vertical cloud overlap over the Tibetan Plateau”, by Jiming Li, Qiaoyi Lv, Bida Jian, Min Zhang, Chuanfeng Zhao, Qiang Fu, Kazuaki Kawamoto, and Hua Zhang.|
The authors present a new parameterization for the cloud overlap statistics over the Tibetan Plateau. The study is well-motivated, providing context for the Tibetan Plateau as a major driver of the Asian monsoon system and the global circulation. Cloud representation biases in climate models lead to errors in the radiation balance over the Tibetan Plateau. The authors use several years of CloudSat data to develop a new cloud overlap parameterization scheme specific to the Tibetan Plateau that incorporates both the vertical wind shear and the atmospheric stability. The scheme compares well against previously developed schemes (for global models) and against the observed cloud overlap.
The paper is well-structured with clearly presented figures. While it is generally well-written, some sentences require grammatical improvement, although this does not appear to lead to ambiguous statements. The science and methodology are sound, and the conclusions are supported by the evidence presented. There are a few questions regarding the methodology, but these should not be considered major issues. I therefore recommend this manuscript to be accepted with minor revisions.
1. Distance between layers
It is not entirely clear how cloud populations are separated by distance between layers. If a continuous cloud layer stretches over 6 km in depth, does it contribute to all distances from 0-6km, or only the maximum distance? If all distances, does it therefore contribute to the “1 km distance” multiple times? That is, in a 6-km deep cloud, we can identify 6 pairs of layers that are separated by 1 km. This issue should be addressed when describing the methodology (see also comment about Figure 1, line 213-221).
The second issue is that multiple cloud types – or clouds at different heights in the atmosphere – may be grouped together. For instance, a contiguous cloud layer that is 1 km deep can occur anywhere, from stratocumulus, to altostratus, to cirrus. I would expect the shear and stability calculated over 1 km to differ a lot for these different cloud types. The authors should address this, perhaps in a brief discussion section.
2. Calculation of cloud cover using parameterization schemes
Section 3.2 is very interesting, but it is sprung upon the reader. The inverse exponential function is not previously introduced. The authors do not explain how the cloud cover is calculated from the different parameterization schemes. Presumably, the decorrelation length scale L is calculated from the dV/dz and dtheta/dz derived from ERA-Interim data interpolated to the CloudSat track. Subsequently, alpha can be calculated for each separation using equation 6. But how does this lead to a calculation of cloud cover (which is compared in Figure 6, according to line 450)? One additional paragraph in this section (prior to presentation of Figure 6) describing the cloud cover calculation is required.
27: One sentence in the abstract on the importance of understanding cloud overlap in the Tibetan Plateau would draw in a broader audience.
32: “Unique” suggests that the authors have compared the TP to all other regions. Perhaps remove this sentence.
68: “Kang et al summarized” – based on what? Observations? Models?
82 & 85: Please specify what type of models you are mostly concerned with, e.g. “horizontal grid length greater than 10 km”. So that it is clear to the reader that only those models rely on some overlap parameterization.
106: Are there any radar sites in the TP region at all?
113: Remove “fortunately”.
124: Break paragraph at “However, the related question…” and merge the remainder with the following paragraph. That sentence is a clear purpose of the paper.
148: Remove “can”.
160: At this point, it is important to clarify that the radar does not distinguish between cloud and precipitation. Then the text at 320-332 will not be such a surprise.
165: Add here the lines 196-199 regarding noise in the observations and surface contamination.
213-221 & Figure 1: This is rather difficult to follow. It would be useful to have one or two additional panels that illustrate the cloud cover and overlap parameter for the particular scene, perhaps for the different length scales considered. For instance, it would be great to have an example of a continuous cloud that is more than 4 km deep, so that the reader can see how it may have “less than random” overlap.
217: What does “cloudy” mean in this context?
233: Please change all references to “discontinuous” to “non-continuous”, or adjust your figures (be consistent).
248: “correlation” – no correlation has been shown or calculated.
253-254: Remove “provided … error. Simply these authors”
260-261: “should account for the typical cloud system scales” add “in their parameterization schemes” (presumably).
281: “the number of available samples” – What is a sample? Is it a 50-km stretch in a CloudSat orbit?
281-294: Figure 2d is rather difficult to interpret. It is likely that these values make more sense when they are presented in a table.
281-294: The description of cloud cover (versus cloud fraction) would have been helpful sooner, probably around line 213-221 (possibly in combination with an illustration in Figure 1).
301: “thicker than other seasons” – is there a simply explanation for this, e.g. a greater tropopause height?
321: “small horizontal scale of cumulus” – The authors should also comment on the fact that these are poorly observed by CloudSat alone, so require the lidar to be available (not extinguished by cloud aloft). How does that affect the statistics?
395: Regarding Figure 5, please mention which scale is used for the segments, presumably 50 km?
400: “seems relatively weaker” – this is difficult to quantify when the two parameters have different units.
411: “As we know” – actually, this is completely new to the reader! (remove)
481: “this new scheme” – please refer to the name of the scheme.
Figure 1: Apart from the comment above (213-221), mention in the caption that observations near the surface have been removed.
Figure 2: What is the uncertainty on alpha? Although the authors provide the sample number in panel d, there could still be a lot of variation in alpha. The authors should provide some measure of uncertainty, e.g. standard deviation or interquartile range.
Figure 5: Again, what is the uncertainty in alpha? The sample number will be smaller due to the compositing on dV/dz and dtheta/dz. The reference to “50% continuous” is confusing in the legend and should be placed in the caption.