|I can see that the authors have spent significant effort on revising the manuscript, which helped improve the manuscript. The largest improvement is a much better statement of the problem (Figure 1), which makes it possible for the reader to understand the motivation and focus of this study without having to read another paper. To me, the problem being studied is the well-known parallax effect in satellite remote sensing although it is given a new name “solar smearing” effect. |
Unfortunately, the fundamental issues I listed in the first round of review still remain and, in my view, the overall quality of the study/manuscript is still far from acceptable for an AMT paper. So, my recommendations remain as “rejection”.
• Problems with the “Emulator”: Although more details are added about their “Emulator” in the revised manuscript, the fundamental problems with it still remain. The authors argue that they can simply “emulate” the impacts of 3-D radiative transfer effects and the parallax effect on TCWV retrieval without using the advanced 3-D radiative transfer model. I simply do not see how this is possible. The author can certainly derive a PIWV based on Eq. (4) and a LES filed, but that is not what the instrument observes (which is radiance or reflectance). A plane-parallel RT model (MODTRAN) is used to “generate the true forward radiance spectra” (page 6 line 9) in this study. But isn’t the “true radiance spectra” the observed spectra that are affected by the 3-D radiative transfer effects and the parallax effect? How could a 1-D model generate the “true” observation? It is mentioned that “in this case, TCWVeff=TCWV” (page 6 line 12). To have TCWVeff=TCWV, we need to have Eq. (6) = Eq. (7) (BTW, Eq. (7) is wrong), which implies PIWV=PIWV_uniform in these simulations. But isn’t the whole idea here is to simulate an ununiform PIWV? In summary, to me there are fundamental problems to the methodology of this study (emulator) which make the results based on it highly skeptical.
• As I mentioned in the first round of review, this paper only presents a phenomenological study of the simulation results, it provides little, if any, explanation of the underlying physics. I think this is due to the use of “emulator” which cannot provide any meaningful explanation of the simulation. Without a solid physical interpretation, these results are highly skeptical to me.
• In section 3.3, the authors found significant differences between the “parallel” and “perpendicular” direction retrievals. Again, there is no explanation of why. Nevertheless, the lack of physical explanation does not prevent the authors from proposing a new sampling scheme to all the possible sensors. To me this result is skeptical to say the least. Any point on the surface can be considered to be either “parallel” or “perpendicular” (or both parallel and perpendicular) to sunlight direction, isn’t it? What is the fundamental difference between any two points in terms of geometry?
In summary, this is a study based on problematic methodology. It provides little, if any, physical interpretation of the simulation results or process-level understanding. I simply do not see why it should be published.