An Improved Acceleration Approach by Utilizing K-Band Range Rate Observations
Jennifer Bonin
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis paper explores an improved acceleration method that utilizes the inter-satellite range-rate observations to estimate the gravity field. There are some points that need to be further considered.
1. In the final paragraph of the introduction, it would be beneficial to highlight the contribution and novelty of the research and elaborate on the specific differences from previous studies more clearly.
2. When introducing the research background, the latest research work done by current scholars is lacking, and the differences between the work done in this paper and other research results are not clarified.
3. Paper format should be further improved, such as Line 109, Line 121…
4. In Figure 2, 3 and 10, when comparing several methods, it is better to mark or enlarge the obvious differences to increase the readability of the figures
Comments on the Quality of English Language-
English expression can be improved, long sentences are difficult to be understood for readers
Author Response
Thank you for your valuable comments. In accordance with your suggestions, we have provided a point-by-point response. Please see the attachment for further details.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsDear Editor and Authors,
The comments are included in the file attachment.
Best regards
Comments for author File: Comments.pdf
Author Response
Thank you for your valuable comments. In accordance with your suggestions, we have provided a point-by-point response. Please see the attachment for further details.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThis paper is yet another which attempts to alter the GRACE processing technique in a small way, resulting in slightly altered GRACE spherical harmonic solutions over the very best five years of GRACE data (and ignoring all the hard parts). The authors have presumably done their math right and get reasonable answers, and I’m sure it was a lot of hard work to set up, which I respect. But ultimately, the results they get are functionally equivalent to the main GRACE data processing centers once a standard smoothing is applied – and they only exist over the very, very best timespan of GRACE. The authors also only test the most basic, spatially-huge regions of the world, omitting any tests of small/narrow regions where differences between various GRACE series can actually be seen.
To make this paper worthy of publication, I recommend these additions/alterations:
1.) Run your proceedure over the ENTIRE GRACE record. That includes the early lousy data, the deep repeats, and the late-year weaker data with gaps. If your solution is similar to CSR/JPL/GFZ during 2005-2010, but falls apart during these hard times, then it is not useful. You admit (P2 L62) that amplification of noise is a known problem of your method, which makes these noisy months an especial concern. This is absolutely key. I would be happier accepting a paper proving that their method does NOT work in hard-to-estimate months, than one which avoids testing them entirely. This is the main reason I am rejecting this paper.
2.) Show how much KBRR data you omit with your outlier reduction method, and give us some evidence that you’re not tossing data in specific geographically-correlated areas. The main solution centers have reduced or eliminated their automatic outlier removal precisely because it was found to eliminate data over real geophysical extreme events. Are you doing the same? That will reduce variance, but not for any good reason!
3.) It is unclear throughout whether you’re really reducing noise, or just reducing ALL signal at the higher deg/ords. I believe that your signal has less high-deg variability. But is that good or bad? To get some sense of this, try averages over SMALL regions (not the huge annual signal in the enormous Amazon!) and narrow ones. For example, can you better separate the coastal mass loss in Alaska using your signal, because you can effectively see to higher deg/ord? Can you separate the sub-basins of the Mississippi? Can you better localize the complex mass changes over the Himalayas? Right now, I doubt it. I think your method just damps ALL signal/noise. But I could be wrong, and I encourage you to prove it. My point is, if you’re not USING the deg > 40 portion of your signal, it doesn’t matter if you did a better job reducing noise in that part of the spectrum. If you can prove that your solution needs less smoothing to reach the same SNR, that’s powerful.
4.) Eliminate all images of the 100km-smoothed versions, and plot the 300km-smoothed instead. Because a glance at Fig 3 shows that your 100km-smoothed is functionally unusable for the vast majority of purposes. Your 300km-smoothed versions (Fig 9) are a level which people will actually USE. Also, add a 500km-smoothed line in all of your stats tables, because that’s what’s usually used over the ocean. (300km is NOT considered an especially “strong” filtering!)
In case the authors decide to revise their paper and resubmit, I offer these smaller suggestions as well:
1.) P2 L52: Please note which major centers use which estimation method.
2.) P2 L87: Why haven’t previous similar attempts used KBRR data? What are the pluses and minuses to using KBRR over KBR using this type of solution? Explain further, for those of us who’ve done other types of KBRR-based estimates. (I had no idea that anyone was NOT using the KBRR data!)
3.) Section 2: This gets rather long and, frankly, I didn’t find it very helpful. For anyone who hasn’t done the “acceleration method”, you skip too many steps to comprehend the details quickly. Anyone who has used the method probably won’t need the math anymore either. I’m sure you did the math right – I don’t need to delve into each line to grasp the basics of what you’re doing. But I’d suggest that a dedicated appendix would be a better place for this – possibly with more intermediate steps and further explanation added, so people can more easily repeat your work. The rest of us will just trust you, frankly.
4.) Equation 17: On the other hand, this is your final result... and I have no real idea what it means. What are C, Q, l, and especially y? Do any of them have a physical meaning? What do the subscripts A, B, rho mean?? What are you ultimately solving FOR? It looks like you’re saying that Grace-A and Grace-B are mathematically unconnected, which seems confusing. Basically, you need to explain your final result more, so those of us who didn’t/couldn’t follow the details can understand how to use what you figured out, at least. Also, Why did you choose a maxdeg of 96?
5.) Please use constant colors for the various solutions in all plots. (Ie: if CSR is red in this plot, make it red in all plots.) Also, if you use Fig 1’s colors, I’d recommend making your own solution black, since that will stand out better from all the older comparison series.
6.) Fig 1: You need to also/instead show an AVERAGE of the degree variance plots (with RMS bars or spaghetti lines on top to show the variance). Because I can’t tell if you’re cherry-picking or these two months are standard.
7.) Fig 2: It is unclear what this depicts. A single month? The stdev of all years?
8.) Fig 3: Please show this with 300km smoothing.
9.) 100km is justly a “mild” filtering, but 300km is only “moderate”. I’d save “strong” for 500km, as that’s very common in oceanographic use. (And please note, I’ve seen people use 1000km filterings for large-scale features before!)
10.) P10 L313: You remove a 6-parameter fit for this case. Do you do that throughout? Make it clear when you do and do not remove the annual/trends.
11.) Fig 4: Please give us statistics (on the figures, in the text, or in a table) listing the mean value of each of the cases here.
12.) Table 2: Make it clear (in both text and caption) that this refers to the stats shown in Figs 5-8, NOT that of Fig 4.
13.) Fig 9: It’s important to note that, after a normal amount of smoothing, your solution has about the same visible level of stripes as every other solution shown.
14.) Fig 10: NO. You can NOT use a color scale which fails to allow differentiation between any of the series. The big red spots all exceed that color scale – and they’re what will show us if you’re over-damping.
15.) Similarly, if you fit/remove the annual signal and trend, and then apply a ~6-month window to knock out the really high-frequency “jitter”, what is the RMS residual of each series? Ie: are you seeing the interannual and 6-11-month signals, or damping those out?
16.) Fig 11: Seriously, you need to plot regions we weren’t looking at way back as examples in 2004. No one is impressed that you can estimate the annual amplitude of the Amazon. Literally every GRACE series in two decades has done that. Look at the HARDER places, where the three main centers’ signals start to vary. That’s where we’ll be able to see if yours is really better, worse, or the same. All I can tell from this is that you didn’t totally screw things up. But there’s no reason to use your product over anyone else’s, in these huge areas.
17.) Table 3 and text: Please don’t crow about getting a near 0.999 correlation with CSR/JPL/GFZ when looking at the PURELY ANNUAL signal in four huge areas! Of COURSE you did – you chopped out any non-annual variability and correlations don’t show variance changes. This statistic is meaningless.
I encourage the authors to redo their work over the full GRACE (and preferably GRACE-FO) record, redo their stats over more challenging basins, and then resubmit.
Author Response
Thank you for your valuable comments. In accordance with your suggestions, we have provided a point-by-point response. Please see the attachment for further details.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsAll comments raised have been addressed appropriately
Author Response
Reply to the reviewers’ comments on the manuscript #2632831 called “An Improved Acceleration Approach by Utilizing K-band Range-rate Observations”
Reply to reviewer#1:
Reply: We express my utmost gratitude for your valuable suggestions.
Reviewer 3 Report
Comments and Suggestions for AuthorsReview 2 of Shen 2023:
I thank the authors for their thoughtful alterations. The current version is vastly more persuasive, particularly thanks to their inclusion of more “iffy” months and comparison over smaller regions. I appreciate the hard work put in and am happy to recommend the current version for publication.
I found two very minor errors/typos in the document that you may want to update:
1.) In section 5, you discuss “four models”. Do you mean the four centers’ solutions (yours, CSR, JPL, and GFZ)? If so, you probably want to use the term “centers” or “solutions” or something like that. The term “model” immediately makes one think of, for example, an ocean or hydrology model being used as a test case.
2.) Fig 11-12: You accidentally left a “m” (meters) on as a label, when SNR is obviously dimensionless. (Oops!)
Author Response
Reply to the reviewers’ comments on the manuscript #2632831 called “An Improved Acceleration Approach by Utilizing K-band Range-rate Observations”
Reply to reviewer#3:
General comments: I thank the authors for their thoughtful alterations. The current version is vastly more persuasive, particularly thanks to their inclusion of more “iffy” months and comparison over smaller regions. I appreciate the hard work put in and am happy to recommend the current version for publication.
Reply: Thank you so much for your very constructive comments.
Comment1: In section 5, you discuss “four models”. Do you mean the four centers’ solutions (yours, CSR, JPL, and GFZ)? If so, you probably want to use the term “centers” or “solutions” or something like that. The term “model” immediately makes one think of, for example, an ocean or hydrology model being used as a test case.
Reply1: Thank you for your valuable comments. Based on your comments, We have made revisions to the corresponding questions in the paper. For more details, please check the revised version.
Comment 2: Fig 11-12: You accidentally left a “m” (meters) on as a label, when SNR is obviously dimensionless. (Oops!)
Reply2: We appreciate your identification of the issue in our article. Accordingly, we have implemented appropriate revisions within the manuscript. For more details, please review the revised manuscript.
