Application and Analysis of XCO2 Data from OCO Satellite Using a Synthetic DINEOF–BME Spatiotemporal Interpolation Framework
Round 1
Reviewer 1 Report
Although the manuscript has been substantially revised, the revisions did not address eleven (that is, almost half) of the comments I labeled “Major issues” in my original review. As the journal does not seem to require point-by-point responses to the comments that contributed to the rejection of the original manuscript, I cannot tell why so many of my concerns were not addressed. Thus, I recommend rejecting the revised manuscript, encouraging the authors to address the issues not addressed in the first revisions, and resubmit a new version of the manuscript along with a point-by-point response to all of my comments (perhaps as part of a cover letter). Please find below the list of comments regarding the original manuscript that, I believe, were not addressed during the revisions—as well as a few comments regarding newly added portions of the manuscript.
---------------------------------
Comments in my original review that were not addressed during the revisions (please note that the line numbers refer to the original manuscript, not the resubmitted one):
1.
Section 2.1: It should be specified exactly which product files were used. This is needed because the links provided in the Data Availability section (Lines 479 and 481) lead to a list of OCO product files and it is not clear, for example, whether this study used the Lite or the Standard product files, and whether or not the study used the “retrospective” datasets. The name of the parameter used as XCO2 data should also be specified. If uncertainty values were also used (given that this is a potential benefit of the proposed method, as discussed in the conclusions section), the name of the parameter providing the uncertainty values should also be specified.
2.
Line 186: It is not clear what is meant by “non-single”.
3.
Line 215: The text shows mu and g at the same level, whereas in Eqs. (4-5) g is higher than mu (as if mu was raised to the power g). This inconsistency should be fixed somehow.
4.
Lines 192-206: These lines are unclear. For example, what specifically is meant by “knowledge base”? It would help if Lines 192-193 were expanded to explain what specific data, spatiotemporal covariance models, physical laws, or scientific theories were part of the knowledge base used in this study. Also, it would be important to clarify what each stage accomplishes. This is needed because, for example, the description of the first stage mentions what the stage considers (I guess the “input”), but it does not tell what this stage produces (I guess the “output”) and how it accomplishes its task.
5.
Line 273: The caption should explain what the line graphs (below and to the right of each monthly) panel show. (My guess is that they show meridional and zonal averages of the interpolated XCO2 concentration values, in ppm.)
6.
Line 279: The caption for Figure 4 seems to be cut off mid-sentence. It should also be clarified whether the EOF values plotted in the left panels are for a specific location or for the overall scene average of the entire study area, etc. It should also be explained what the colors represent in the right-side panels and in what units. (The text mentions that the panels show the “corresponding parameter representing the variability of the original signal explained by each EOF mode”, but does not say what this “corresponding parameter” and its units are. My guess would be that it is deviation from the annual mean XCO2 concentration averaged over the entire scene, in units of ppm—but I am not confident in this guess.)
7.
Line 299 and Tables 1 and 2: It would be important to calculate and report not only these three error statistics, but also the perhaps most basic one, the bias (i.e., mean error). If appropriate, these bias values could also be included into the conclusions and perhaps even into the abstract.
8.
Also, similarly to Table 1, Table 2 should also include the values for the stand-alone DINEOF method (not just for DINEOF-BME and OCO). This would probably require splitting the table into several segments, as all the new columns may not fit next to the old ones (especially when the columns for mean errors are also added, as recommended above), but this shouldn’t be a problem for the journal.
9.
Line 360: It should be explained why the DINEOF method cannot achieve 100% coverage. At first I thought that DINEOF provided invalid values in some unusual situations—but then I realized that this cannot be the case, as the coverage is constant at 97% even though the individual interpolation situation is very different in each month.
10.
Lines 461-462: It should be clarified in what sense the inclusiveness of the BME method got demonstrated. Did this particular study use some specific pieces of data that other methods could not incorporate?
11.
Line 484: This link does not work: When I click on it, I get the error message “Oops, that page can’t be found.”
--------------------------------------
Comments about newly added portions of the manuscript (with line numbers referring to the resubmitted version):
1.
Lines 197-198: It should be explained/clarified how the matrix “I” is used in the method. As is, the matrix is defined but it does not seem to be used. (It was used in the original manuscript but not in the resubmitted one.)
2.
Lines 443-476 contain an exact copy of Lines 411-442 and should be deleted.
Author Response
Please see attached file
Author Response File: 
Author Response.docx
Reviewer 2 Report
Dear Authors,
The paper "Application and Analysis of XCO2 Data from OCO Satellite using a synthetic DINEOF-BME spatiotemporal interpolation framework" presents a statistical approach in the use of scattered spatial data that changes over time. This approach is an alternative to more sophisticated data assimilation in the model of the general circulation of the atmosphere and the transport of trace gases. In this regard, the paper would be of great interest to readers if it were written properly. To my deep regret, this text needs further improvement in order to become understandable to the reader.
In my opinion, this study cannot be published without the following changes:
1. It should be explained to the reader how the essentially two-dimensional (longitude and latitude) data shown in Figure 1 is ordered into a single dimension (row number) in the data matrix X in Eq. 1.
2. Formulas that include matrices should be rewritten using the generally accepted notation, and also explained in the text, where it is supposed to multiply matrices according to the rules of matrix multiplication, and where component-by-component multiplication is supposed. For example, matrices should be denoted by bold capital letters, with no indexes for the size of the matrices. And if the components of the matrix with indices are meant, then they should be in small letters of regular thickness. If Eq. 2 is a matrix formula, then it is incorrect, since the number of rows in the first matrix is not equal to the number of columns in the second.
I apologize for the metaphor, but there is a persistent feeling when reading section 3.2 that one researcher did the work, and another wrote the text of the section, who did not quite understand what the first one was doing.
After the reader stumbles upon the two problems mentioned above, further reading of the article loses its meaning for him.
Some improvement in language and style of presentation is also needed. For example, the abstract has the first use of an abbreviation DINEOF. You should write the exact transcript of this abbreviation, word by word.
Sincerely yours,
Your Reviewer.
Author Response
Please see attached file.
Author Response File: Author Response.docx
Reviewer 3 Report
The paper combines the empirical orthogonal function interpolation (DINEOF) and the Bayesian Maximum Entropy (BME) methods to interpolate assimilated XCO2 data of Orbiting Carbon Observatory 2(OCO-2) and Orbiting Carbon Observatory 3(OCO-3) data in the time range January to December 2020 within the geographical range of 15-17 45°N and 120-150°E. This application allowed to reach a fine resolution of satellite data compared with the previously used methods. The results are validated on the in-situ measured XCO2 data provided by the Total Carbon Column Observing Network (TCCON) to demonstrate the efficiency of the applied method of interpolation.
My only concern is a limitation of this application to a relatively small geographical region (15-45 N, 120-150E) on monthly time scale. It would be more beneficial for the climate community to see the expansion to global data an the shorter, say daily, time scale. I recommend this to authors for future work. However the authors achieved the important goal of the demonstration of their method and I recommend the paper to publication in Remote Sensing journal.
Author Response
Please see attached file.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The authors addressed most of my comments appropriately. I still recommend some changes, but the manuscript is getting close to being publishable. Please find my specific comments below.
1.
My original comment:
Comment 2. Line 186: It is not clear what is meant by “non-single”.
Response by the authors:
What we want to express here is that this study can not only use uncertain data to assist interpolation, but also fuse multiple uncertain data, and adjust the soft data type according to the type of uncertain data.
My comment about the authors’ response:
Thanks for the explanation. In this case the wording “uncertain non-single data information” should be changed to something like “uncertain data from multiple sources” or “multiple types of data and uncertainty values” or something similar.
2.
My original comment:
Comment 4. Lines 192-206: These lines are unclear. For example, what specifically is meant by “knowledge base”? It would help if Lines 192-193 were expanded to explain what specific data, spatiotemporal covariance models, physical laws, or scientific theories were part of the knowledge base used in this study. Also, it would be important to clarify what each stage accomplishes. This is needed because, for example, the description of the first stage mentions what the stage considers (I guess the “input”), but it does not tell what this stage produces (I guess the “output”) and how it accomplishes its task.
Response by the authors:
About the ‘knowledge base’ used in this study has been added.(Lines 240-251)
In BME method , the ‘knowledge base’ include ‘general knowledge’ and ‘site-specific knowledge’. The ‘general knowledge’ in this study is covariance function.(Lines 240-244).The site-specific knowledge consisting of hard data and soft data. (Lines 246-251)
The output of each process is illustrated in Figure 2. The input stage refers to the prior stage and the intermediate stage, while the output stage refers to the posterior stage. The prediction value of studied natural attribute (here is concentration of XCO2) can be calculated by the posterior PDF obtained in the posterior stage.
My comment about the authors’ response:
The meaning of the acronym “STRF” in Line 243 should be explained. On the other hand, the abbreviation “RV” defined in Line 243 does not seem to be used anywhere and could be deleted. Also, in Line 242 the reference is to a single Equation (10), and so “Eqs.” should be changed to “Eq.”.
Figure 2 is indeed of some help, but it would still be an important further help if the various items in the figure with generic names such as “Private data” would be explained with in the figure or in the manuscript text. For example, it should be clarified that “Private data” means site specific data, and ideally even the list of parameters included in the “Private data” category would be provided.
3.
My original comment:
Comment 6. Line 279: The caption for Figure 4 seems to be cut off mid-sentence. It should also be clarified whether the EOF values plotted in the left panels are for a specific location or for the overall scene average of the entire study area, etc. It should also be explained what the colors represent in the right-side panels and in what units. (The text mentions that the panels show the “corresponding parameter representing the variability of the original signal explained by each EOF mode”, but does not say what this “corresponding parameter” and its units are. My guess would be that it is deviation from the annual mean XCO2 concentration averaged over the entire scene, in units of ppm—but I am not confident in this guess.)
Response by the authors:
Thank you for your suggestion, and the explanation about the missing part in the figure has been added as follows(Lines 345-351).
“The value of EOF modal parameter represents its variance contribution rate, where the line chart of time series represents the change of the whole study area over time, and the spatial modal chart represents the region that conforms to the change of time series. If the spatial modal coefficient is regular, it means that the change of time series in this space conforms to the corresponding time mode. On the contrary, if the spatial modal coefficient is negative, it means that the time series change of this space is opposite to the corresponding time mode.”
My comment about the authors’ response:
The word “regular” in Line 348 should be replaced by “positive”, or the meaning of “regular” in this context should be explained.
4.
My original comment:
Comment 9. Line 360: It should be explained why the DINEOF method cannot achieve 100% coverage. At first I thought that DINEOF provided invalid values in some unusual situations—but then I realized that this cannot be the case, as the coverage is constant at 97% even though the individual interpolation situation is very different in each month.
Response by the authors:
Because DINEOF method uniformly calculates the values of all time units at the same space point, the results obtained by DINEOF method will have the same coverage.
The DINEOF method is a purely mathematical method. Just as Eq.9 when it calculate the prediction value of one spatial-temporal point would use all value of points that in the same time unit or space unit .Therefore, in this study, when all time units of a space point are vacant values, all time units of the space point can not be predicted, so the coverage of results obtained in each time unit is the same. Due to the fact that the coverage rate of XCO2 is extremely low, even 3% of the spatial points don’t have valid values in a continuous 12-month temporal span, the DINEOF method cannot give effective perdition at these kind of points,.
My comment about the authors’ response:
I recommend including a sentence about this into the manuscript itself, for example something along the lines of “The coverage of the DINEOF method does not reach 100% because this method cannot provide output for the 3% of cases where no data is available at sufficiently close time and locations for the spatiotemporal interpolation.”
5.
My original comment:
Comment 11. Line 484: This link does not work: When I click on it, I get the error message “Oops, that page can’t be found.”
Response by the authors:
We have tested all the links in the article to ensure that all the links in the current article are clickable and successful jump.
My comment about the authors’ response:
This might be a problem only for the marked-up version that highlights the most recent changes made to the manuscript: The first two updated links do not work when I click on them, as my browsed tries to go to a URL that includes both the updated and the original URLs. (For example, to https://oco2.gesdisc.eosdis.nasa.gov/data/OCO2_DATA/OCO2_L2_Lite_FP.10r/https://oco2.gesdis )
Author Response
Dear reviewer:
Thank you very much for your reply. I have replied to your suggestions in the attached file. Please see the attachment.
Yours,
Yutong Jiang
Author Response File: Author Response.docx
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
In the manuscript titled " Application and Analysis of XCO2 Data from OCO Satellite using a synthetic DINEOF-BME spatiotemporal interpolation framework" , the Authors combined DINEOF and BME to map XCO2 in the composite space-time domain. The idea is good, but currently, this research is not conducted correctly and has several flaws.
The cross validation of DINEOF-BME approach seemed to have higher accuracy, however, its comparative analysis with SOCAT data is not satisfactory. The current global average of XCO2 is about 400 ppm. If the error of the comparative analysis has reached almost 50 ppm, I donot think such a comparative analysis is meaningful. It is recommended that the authors use other comparative data to compare and analyze DINEOF-BME interpolation results.
Besides, English in this manuscript seems relatively poor. Several lines/paragraphs are confusing and many spelling mistakes. XCO2 should be "XCO2", pay attention to the subscript. Authors should consider rephrasing and modifying.
Reviewer 2 Report
This study examines the potential of combining two approaches for a joint spatiotemporal interpolation of satellite-based carbon-dioxide measurements. The goal is to obtain continuous fields of carbon-dioxide concentrations at a high resolution and wide coverage. Unfortunately, in my opinion, the paper needs substantial changes to make it suitable for publication. Although I did not find fatal flaws in the study, the sheer number of major issues raised in my review prevent me from recommending the publication of the manuscript. Instead, I recommend rejecting the current manuscript and encouraging the authors to consider revising and resubmitting it. I recommend rejection both because the revisions would likely require more time than typically allowed by journal policies and because, given the amount of currently missing or unclear information, I cannot tell with confidence whether or not clarifying and fixing the issues raised in my review would make the paper suitable for publication. Please find my specific comments below.
Major issues:
Section 2.1: It should be specified exactly which product files were used. This is needed because the links provided in the Data Availability section (Lines 479 and 481) lead to a list of OCO product files and it is not clear, for example, whether this study used the Lite or the Standard product files, and whether or not the study used the “retrospective” datasets. The name of the parameter used as XCO2 data should also be specified. If uncertainty values were also used (given that this is a potential benefit of the proposed method, as discussed in the conclusions section), the name of the parameter providing the uncertainty values should also be specified.
Lines 145-146: The text “S is the dimension of data X in space and T is the dimension of data X in time” is not quite clear. For example, it would help to clarify whether, in this particular study, S is the number of 0.1 degree by 0.1 degree areas within the study area (S=30X30/(0.1*0.1) = 90000) and T is the number of considered months (T=12).
Line 149: It should be clarified what “rho”, “N” “K”, and “i” stand for. Also, it should be explained what “U” and “V” are (the text only mentions their size, but not what they are).
Line 158: It would help to clarify whether “i” indicates location and “j” identifies time.
Line 163: The wording “the mean value on the time dimension” is not clear. Does it refer to the mean value of the valid XCO2 data in a certain month? Also, should the word “from” perhaps be moved to the beginning of the line (to right after “subtracting”)?
Line 186: It is not clear what is meant by “non-single”.
Line 215: The text shows mu and g at the same level, whereas in Eqs. (4-5) g is higher than mu (as if mu was raised to the power g). This inconsistency should be fixed somehow.
Lines 192-206: These lines are unclear. For example, what specifically is meant by “knowledge base”? It would help if Lines 192-193 were expanded to explain what specific data, spatiotemporal covariance models, physical laws, or scientific theories were part of the knowledge base used in this study. Also, it would be important to clarify what each stage accomplishes. This is needed because, for example, the description of the first stage mentions what the stage considers (I guess the “input”), but it does not tell what this stage produces (I guess the “output”) and how it accomplishes its task.
Lines 206-217 and Lines 192-206: My recommendation is either to delete Lines 206-217 (as they cannot be understood if one is not already familiar with the BME method) or to insert a more detailed description. In the latter case, the description should provide not only names for each term in Eqs. (4-5), but should explain their physical/mathematical meaning or purpose, and how each term may be obtained. For example, what does the normalization factor A normalize to and why, what law is referred to as the “XCO2 probability law fk”, etc. In this case it should also be explained how the three stages described in Lines 191-202 are represented in equations (4-5): for example, are some terms representing operations of the first, second, or third stage or perhaps the inputs/outputs of these stages?
Lines 254-257: This sentence is not quite clear and may slow down readers significantly. (My guess is that, in essence, we check whether using 80% of the data allows us to accurately estimate the remaining 20% of available data values—and then we repeat this process by checking the results if we select a different one-fifth of the data as the “target”.)
Line 273: The caption should explain what the line graphs (below and to the right of each monthly) panel show. (My guess is that they show meridional and zonal averages of the interpolated XCO2 concentration values, in ppm.)
Line 279: The caption for Figure 4 seems to be cut off mid-sentence. Also, the figure should indicate what is displayed along the horizontal axis (my guess is “month of the year”), perhaps by adding an axis title to the figure panels. It should also be clarified whether the EOF values plotted in the left panels are for a specific location or for the overall scene average of the entire study area, etc. It should also be explained what the colors represent in the right-side panels and in what units. (The text mentions that the panels show the “corresponding parameter representing the variability of the original signal explained by each EOF mode”, but does not say what this “corresponding parameter” and its units are. My guess would be that it is deviation from the annual mean XCO2 concentration averaged over the entire scene, in units of ppm—but I am not confident in this guess.)
Lines 289 and 290: The words “frequent” and “frequency” don’t seem to fit. Perhaps something like “intensity” could work better. However, the entire sentence in lines 290-291 seem problematic. I would guess that as we go farther away from land, XCO2 will decrease mainly not because the local human activities will be weaker in the middle of the ocean than near the shore (with fewer boats farther offshore), but because the impact of distant land-based XCO2 emissions caused by human activities will be weaker as the distance from land increases. (In other words, wind-driven XCO2 transport from land will bring less XCO2 to maritime areas that are farther away.)
Line 299: I guess “root square error” should be replaced by “mean square error”, the quantity that is shown in Table 1. Also, “mean deviation” should be replaced by “mean absolute error”, the quantity shown in both Tables 1 and 2. After the name of each quantity, the acronyms (MAE, MSE, RMSE) should also be specified in parentheses.
Line 299 and Tables 1 and 2: It would be important to calculate and report not only these three error statistics, but also the perhaps most basic one, the bias (i.e., mean error). If appropriate, these bias values could also be included into the conclusions and perhaps even into the abstract.
Line 317: The exact definition of parameters “count” and “coverage” should be explained. At first I thought that “count” refers to the number of SOCAT observations that could be compared to interpolated or satellite-observed values, and “coverage” tells what percentage of the available SOCAT data could be compared to interpolated or satellite-observed values. However, I later realized that this cannot be right: If the number of SOCAT data for January was 299 (as it would be implied by the 100% coverage for interpolated data), then a single point from OCO could not achieve 5% coverage for this month.
Also, similarly to Table 1, Table 2 should also include the values for the stand-alone DINEOF method (not just for DINEOF-BME and OCO). This would probably require splitting the table into several segments, as all the new columns may not fit next to the old ones (especially when the columns for mean errors are also added, as recommended above), but this shouldn’t be a problem for the journal.
Finally, it should be clarified somewhere around here whether my guess is correct and Table 1 shows statistics of the 12 monthly mean values and that Table 2 shows error statistics for the individual matchups with SOCAT observations. It should also be clarified whether the average value in the bottom row is the average of the 12 monthly values shown above or the overall combined statistics of all individual matchups with SOCAT data throughout the whole year. (A small point about aesthetics: The first two numbers in this “average” row should be nudged upward slightly or the other values should be nudged downward slightly to make all values line up nicely.)
Lines 330-331: The words “largely accidental” should be replaced by something along the line of “highly uncertain”. Also, this sentence raises the question whether the available dataset is sufficiently large to allow a meaningful validation—or perhaps that the small number of OCO-SOCAT matchups is helpful for demonstrating the benefits of the proposed DINEOF-BME method.
Line 338: It should be mentioned somewhere that Figures 5 and 6 display values that were already shown in Table 2 (with the exception of the 97% coverage for the DINEOF method). Perhaps the figures could be deleted, or some features that are easier to spot visually than by looking at a table could be pointed out.
Also, it would be important to discuss somewhere a feature of the results in Table 2 and Figure 5: That in several cases, the validation results indicate higher accuracy for the interpolated data than for the original OCO data on which the interpolation was based. It would help to discuss whether the improvement comes from the temporal aspect of the interpolation or from the uneven spatial sampling of OCO data being mitigated by the uniform, full coverage of the interpolation.
Line 360: It should be explained why the DINEOF method cannot achieve 100% coverage. At first I thought that DINEOF provided invalid values in some unusual situations—but then I realized that this cannot be the case, as the coverage is constant at 97% even though the individual interpolation situation is very different in each month.
Lines 376-390: While the manuscript discusses the importance of satellite-based carbon-dioxide concentration datasets and the need for filling the gaps between available observations, it does not show why there is a need for new methods for this interpolation, and whether the method proposed in this study offer real benefits over existing methods such as the kriging techniques discussed. Do the results suggest that the method is superior or at least competitive when compared to the kriging methods in terms of accuracy or in terms of ease of use or setup? (Some theoretical advantages of the proposed approach are mentioned in the manuscript, but the potential contribution of the paper would be much stronger if it could argue that these theoretical advantages translate to actual practical benefits.)
Lines 388-390: It would help to mention somewhere in the manuscript the typical uncertainties of XCO2 values provided by SOCAT and by the OCO instruments (especially since a key theoretical advantage of the proposed approach is the ability to use the uncertainty of input data values).
Lines 458-459: Since the previous two sentences reveal that the RMSE and MAE errors of the proposed method are slightly larger than the errors of the original satellite measurements, it is not correct that “the MAE and RMSE values of interpolation products have been slightly improved”. As a result, the word “improved” should be replaced by something along the lines of “increased” or “worsened”.
Lines 461-462: It should be clarified in what sense the inclusiveness of the BME method got demonstrated. Did this particular study use some specific pieces of data that other methods could not incorporate?
Line 484: This link does not work: When I click on it, I get the error message “Oops, that page can’t be found.”
Wording issues:
Line 13: It is not clear what the word “them” refers to. Perhaps “others” would fit better?
Lines 15 and 444: DINEOF does not stand for “empirical orthogonal function interpolation”; instead, as explained in Line 69, it stands for “Data Interpolating Empirical Orthogonal Functions”.
Line 21: The word “for” should be deleted.
Line 26: “BME-DINEOF” should be replaced by “DINEOF-BME”.
Lines 43-44: The words “Among them” don’t fit well here either, but here I’d suggest some other replacement; perhaps “Among these factors” could work.
Lines 54-55: The spacing on both sides of the slash should be consistent: either there should be a space on both sides or there should not be a space on either side.
Line 63: The expression “atmospheric sol” is rarely used; “atmospheric aerosol” would be much more widely recognized and understood.
Line 84: I suggest replacing “and” by “but”.
Line 97: I recommend inserting “described” or “presented” between the words “is” and “as”.
Line 98: “Part 2” should be changed to “Part 3”. Also, the letter “n” seems to be missing from “handling”.
Line 117: The acronym “SOCAT” stands for “Surface Ocean CO₂ Atlas”, not for “Ocean Surface Carbon Dioxide Atlas”.
Line 122: I recommend replacing “In this study” by something along the lines of “As it will be described in the next section”.
Line 152: A dot is missing at the end.
Line 153: I recommend deleting the second “the”.
Line 157: I guess a letter “t” should be added at the end of “no”.
Line 170: A letter “s” should be added at the end of “value” and “point”.
Lines 218-220: If the presented study happened to use any of these software libraries, it should be mentioned; otherwise, the sentence could be rephrased, perhaps to say that although this particular study did not use them, several software libraries are available.
Line 245: An “s” should be added either at the end of “EOF” or at the ends of “analyze” and “extract”.
Line 264: The word “on” should be replaced by “at”.
Lines 271 and 272: I recommend changing “is inversely proportional to” to “decreases linearly with”, and “proportional” to “increases linearly with”, because proportional would mean that a twice as large distance would results in twice as large (or half as large) XCO2 values.
Line 274 or Figure 3: The units should be specified either as a label to the color bar or in the figure caption. (My guess is ppm.)
Lines 303-304: The text “in terms of average deviation and RMSE value” should be deleted.
Line 308: I suggest switching the order of the two rows in Table 1 (putting DINEOF above DINEOF-BME).
Line 431: One of the two occurrences of the word “more” should be deleted.
Line 439: The word “are” should be deleted. More importantly, the word “rules” should be replaced by something like “behaviors”.
Reviewer 3 Report
Dear Authors,
Your paper presents a work on the spatial-temporal interpolation of OCO-2 and OCO-3 data on the total carbon dioxide content in the atmospheric column (XCO2) using combined technique of empirical orthogonal functions and Bayesian estimation. The work itself is of interest to the scientific community, since alternatives to the assimilation of satellite sensing data in the model of atmospheric circulation and transport of chemical tracers can be much easier to implement by individual scientific groups.
To my great regret, your great work is presented in the text of the paper in an inappropriate way. The text is very difficult for the reader to perceive, the number of corrections that I would have to recommend to the authors is too large. I will give just some examples of an unsuccessful style of presentation. For example, the expression "carbon satellite" is suitable for oral speech in a conversation with colleagues, but is not suitable for a journal article. In general, the English of the entire paper should be improved. It is enough for authors to simply imagine themselves in the place of a reader who should be treated with respect. The design of formulas and the description of the quantities included in them are also not performed at the proper level.
Another example. A historical digression into the use of BME method in lines 77-83, in my opinion, is not necessary at all. Authors also need to carefully check the correctness of the first use of abbreviations.
Here is an example of another remark. Formula (1) uses a singular value decomposition of the matrix, i.e. an object that can be considered two-dimensional, but satellite XCO2 data can be considered three-dimensional (longitude, latitude, time), therefore, it should be explained to the reader how two-dimensional coordinate data are ordered along one dimension.
Based on the above, my recommendation to the authors is to completely carefully rewrite the article (at least sections 1 and 3 without fail) taking into account the interests of potential readers and submit it to the journal again. In the form in which it is now, it cannot be published.
Sincerely yours,
Your reviewer.
