This review represents my response to points made in the Author's Response uploaded on Oct 8th, primarily concerning issues raised by the reviewers that were in no way addressed by the authors. I recommend the paper for further revision.
Each time I saw a response of “I could not see this comment in Revision 2”, I had a desire to reject this paper completely as the authors appear to be arguing in bad faith. You submitted a major revision and there was over a year between our reviews. We brought fresh eyes the second time. Just because a comment wasn't brought up in the previous round doesn't mean you can ignore it now.
I am aware that this was not the kindest series of reviews. We think that your method is conceptually interesting but see little evidence that it is worthwhile from the evidence you present here or in Part II. You provide no evidence that the use of discrete aerosol classes produces a worse product – merely a conceptual argument that continuous variation is more rigorous. Though I agree with you on a technical level, your critique of existing retrievals uses unduly aggressive language and, to a native speaker, comes across as a personal attack. Your continual refusal to tone down your language implies you genuinely wish to insult the entire field. I shall stop requesting changes there and simply wish you luck.
Your title remains misleading. I have asked several of my colleagues, including those outside of atmospheric science. The phrase “continuous variation of the state variables in solution space” was interpreted to mean “our state vector and prior information were cast in terms of surface reflectance and aerosol properties”. You definitively do not do that for aerosol properties. You defined some representative aerosol models, which you refer to as vertices, and retrieve a linear combination of those, from which the SSA and asymmetry parameter are derived. To be clear – there is nothing wrong with that idea. However, you title should represent the work you did. It should mention that your method uses the linear combination of idealised aerosol models (or words to that effect).
Specific points, in the order I encountered them, follow. Line numbers below refer to those given in the Author's Response.
- I considered the SSA and asymmetry parameter to be microphysical properties but, looking into it, I see that is not a common opinion. My apologies for offering inappropriate terminology for your title.
- “Why decouple the papers when there's only one forward model?” Apologies for my poor phrasing. I meant, “Why have a single forward model to retrieve aerosol and surface reflectance but describe the manner of their retrieval in two independent papers separated by 8 years?” Yes, you didn't change the surface forward model and, yes, surface reflectance is less visible in the current research climate, but I would have liked to see a discussion and presentation of the non-linear interactions of these two coupled variables that are evaluated separately everywhere else. You're experts in both fields – a rare combination; I would have liked to hear what you had to say.
- Your example of a merged Figs. 6-9 is another case where the authors appear to be arguing in bad faith. They merge Figs. 5-12, which is indeed difficult to read. Showing only the three (5, 6-9, and 10-12, making three total images) we asked for will also be messy, but that can be alleviated by (a) setting the error bars to have a transparency (using the alpha keyword in matplotlib) and using additional colours to distinguish the three curves. The Brewer colour table, “tab10” in Python, has 11 colours designed to be distinguishable in different light levels and by people with various sight problems, which should be sufficient. (As this journal is not physical published, except on demand, the generation of solely colour images should not be a concern.)
L20) You described the lower layer of your conceptual radiative transfer scheme as “soil/vegetation strata”, which the reviewer disagreed with. When retrieving surface reflectance, one may use multiple “layers” within the vegetation layer. When retrieving atmospheric properties, the surface is typically a boundary condition on the system (as opposed to an active component). If your retrieval actually has layers in both the atmosphere and surface, keep your current sentence but replace “bottom” with “lower” as the former word implies a singular element. If not, try splitting the sentence and briefly describing the problem in both (i.e. in the atmosphere, layers represent temperature, ozone, the free troposphere vs boundary layer, etc. while near the surface, layers represent the canopy, under-canopy, brush, etc.).
L20) The comment was asking you to say “can be further complicated” rather than “is further complicated”.
L50) I can understand why the reviewer asked to use “models” instead of “classes”, as the latter can describe a collection of the former. You may have difficulty in future work if you maintain the current terminology, but that's your choice to make.
L59) Try “approaches under-perform compared to methods” to reduce the ambiguity here.
L61-65) The reviewer meant that these lines repeat a sentiment you made twice already in the introduction. The second sentence, in particular, seems redundant. However, I see why you wanted to make a summary of the paper's intent here. As an alternative, a brief statement of what you do that is unique (e.g. calculating SSA and g from the linear combination of aerosol models) would be useful for a reader that skips directly to the conclusions.
L90) While VIS0.6 is the official name of the bands, most of your readers aren't going to know that. I note that this disagreement arises repeatedly throughout the paper, so I'm going to assume you want to us the engineering terminology of “bands”. I leave it to the Editor to decide if that is appropriate for the audience of this journal.
L93) “smallest cost” would be more accurate than “best fit” as the latter doesn't have a single, objective meaning.
L100) “a continuous variations” is not valid English. “the continuous variation” would be grammatically correct but, as the reviewers have complained in two successive reviews, isn't likely to be understood by most of your readers. You appear to have no intention of changing it, presumably because the recent ESA ITT requested “continuous variation of the aerosol optical properties” and you have a box to tick.
L101) It would be nice if you said “preferred” rather than “required”.
L103) Your response to the comment here is good. Writing “but uncertainties cannot be assigned to this choice in a straightforward manner” would be useful.
L105) Try “information and its associated uncertainties” to improve clarity.
L116) Probably remove “or planned” as both 3MI and PACE intend to offer multi-angular observations and, to different extents, polarisation.
L119) Replace the comma with “by” and I think you address the reviewer's comment.
L127) You've made the response “No comment were made on this paragraph in revision 2” repeatedly. What, exactly do you mean by this? I've checked the files stored online, and these comments are all present there. If you mean “you didn't complain about it the first time, so why should I fix it now,” that is the height of arrogance. I will give you the benefit of the doubt and assume what you meant is “we don't know why this paragraph is confusing to you.” I empathise that comments like “Please reword” aren't very useful as you will have tried your best to write to be understood. In this case, I would recommend,
Putting a comma between “properties” and “such” on L128.
Confusing sentences can be simplified by moving interrupting statements into the main sentence. An alternative to the sentence starting at L129 is, “The objective of retrievals that assume aerosol classes is to provide a reasonable sampling of the [g, w0] space. Omitting areas of that space may produced biased retrievals, as discussed in Govaerts et al. (2010).”
Try “in this paper” rather than “by these authors” as the latter could be thought to refer to the authors of the 2010 paper.
For the sentence on L133, perhaps, “That choice of classes was intended to provide a sampling of solution space representative of real-world conditions. The inversion is repeated for each aerosol class and the result with the best fit is reported (rather than vary the aerosol properties continuously, as would be preferable).”
L138) The reviewer makes an interesting comment. Looking into it, your use of braces {} to denote the set {g, w0} is accurate. I'm less certain of using square brackets to denote the solution space. Definitions of linear vector spaces appear to use braces (and using {} everywhere would look nicer), but I think this is a matter for the typesetter.
Fig.2) If you're using LaTeX, I will point out the existence of \textmu, which can be used outside math mode for a non-italic character.
L144) Your web reference for these definitions could be a good footnote.
L146) The copy editor will need to add commas to this sentence.
L154) “relatively” would be better than “almost”.
L177) Try “This work” rather than “The present study”.
Fig.4) Add “gas” before “absorption”.
L201) Yes, the gases depend on the wavelength but, if someone is going to replicate your work from this paper (the ideal we aspire to), they need to know which of the major species you accounted for. You don't need to be specific as to where they're used, as that would take too long, but the different aerosol groups vary in what gases they consider and Patadia et al. (2018, doi:10.5194/amt-2018-7) showed that that choice matters.
L207) Move the \Phi to after “parameter” to be clear what it stands for.
L222) How about “in the Fourier space for all illumination and...”?
L225) The reviewer meant you should define the general meaning of the word Jacobian. Some members of our field were not educated in the physical sciences as undergraduates and it's friendly to not assume they know what all our terms mean.
S4.4L1) Perhaps “principal” rather than “most important”?
L237) You don't need “As can be seen”. It's the sort of thing we say out loud to give us time to remember what to say next but just wastes time when written.
L238) “Another” is one word.
L239) The copy editor is going to ask you to define PROBA-V. You can omit the word here if you want to wait to define it till later.
L299) By your notation, its obvious that \Delta \tau_a is a change in \tau_a. Explaining it simply made me stop and think about things to work out what you meant.
L7 after 295) Try “in two successive iterations” rather than “twice consecutively”.
L303) Maybe “chosen” rather than “delineated”.
L303) Add 's after CISAR.
Tab.2) I believe the reviewer's sarcastic remark was attempting to point out that you can give values to X significant figures rather than Y decimal places. However, if you actually used these precise figures, leave them in so that we may all judge your undue precision.
L309) The reviewer appears to have misunderstood what you meant by “assumed that the surface parameters are known a priori”. Perhaps this sentence could read, “In these experiments, to concentrate on the retrieval of aerosol properties, the surface parameters are set to the true values used in simulation (Table 2) and are ascribed a small uncertainty of 0.03 (though they remain part of the retrieved state vector).”
Tab.4) The first “experiment” should be plural and the second singular.
Fig.5) There is always uncertainty. You just might have reason to believe it's really small.
Tab.6) The plus/minus notation is a good idea. The fact it didn't occur until now isn't justification to not do it.
Tab.7) I suspect the typesetter will have opinions about where the % sign should go.
L383) Perhaps “at” rather than “in”.
L385) “The CISAR algorithm retrieves total AODs consistent with the truth.” would be a better sentence.
L387) Technically, the vertex is not a degree of freedom but that is the phrase that would be used in the scientific vernacular. Maybe put quote marks around it?
L389-401) I can understand why the reviewer responded strongly to this paragraph. Intuitively, it feels like adding more variation should produce a better retrieval. What I suspect is happening is that the fourth vertex can be closely mimicked by some combination of the other three vertex together (in the terminology of linear algebra, it is almost linearly dependent on the other three). The increase in uncertainty represents a flattening of the cost function because changing the input of vertex CS is similar to changing the others.
A more detailed discussion of this would be interesting, replacing “is therefore not straightforward”. The word “overconstrained” would likely appear.
L389) I don't believe you can state that the use of *any* two coarse mode vertices is worthless. Perhaps replace “adding two coarse mode vertices” with “using these two coarse mode vertices”. (I'd hope you did some work to eliminate any obvious pairings and, if you did, “using any obvious pair of coarse mode vertices” would be appropriate.)
L392) Further to that point, could you add make this sentence, “This series of experiments has shown that, of the four considered, the use of the FN, FA and CL vertices provides the highest quality retrieval of aerosol class F1.” We wouldn't want someone to cite this paper as proof that only these three vertices are necessary to retrieve any aerosol.
L444) Could you mention that experiments with noise perturbation are included in the supplement? |
- Printer-friendly version
- Supplement
