Application of Optimal Interpolation to Spatially and Temporally Sparse Observations of Aerosol Optical Depth
, Gennadi Milinevsky 2,3,4,*, Andrey Bril 1, Anatoly Chaikovsky 1, Alexander Miskevich 1 and Yuliia Yukhymchuk 2,5
Round 1
Reviewer 1 Report
The paper presents an application of optimal interpolation to aerosol optical depth observations. It is a suitable topic for Atmosphere MDPI journal. The manuscript is professionally written, clear, and easy to read. Furthermore, it is of great interest and importance to the research field. Therefore, I would recommend a revision following my comments below.
- The main objectives and results of the manuscript should be highlighted in the Abstract.
- The first paragraph of Introduction is too long. It could be split in two or three paragraphs.
- The authors should improve the Results section with the discussion of Tables 1 and 2 present in Section 4. And a more elaborate Discussion section should be provided.
- The manuscript's novelty should be (more) explored in the Abstract, Introduction, and Conclusions.
Author Response
Response to Reviewer 1 Comments
The revised manuscript atmosphere-2064524 by Miatselskaya et al. Application of Optimal Interpolation to Spatially and Temporally Sparse Observations of Aerosol Optical Depth
We thank Reviewer 1 for the useful comments and suggestions that have helped to improve our manuscript. Our responses to the specific points are included below in blue.
Reviewer 1 Comments and Suggestions for Authors
Minor English changes required
Authors Reply (AR). We improved our English where possible and used Premium Grammarly as well.
Reviewer comment (RC).
The paper presents an application of optimal interpolation to aerosol optical depth observations. It is a suitable topic for Atmosphere MDPI journal. The manuscript is professionally written, clear, and easy to read. Furthermore, it is of great interest and importance to the research field. Therefore, I would recommend a revision following my comments below.
RC: The main objectives and results of the manuscript should be highlighted in the Abstract.
AR: Corrected, abstract re-written, the main objectives and results of the manuscript are highlighted.
RC: The first paragraph of Introduction is too long. It could be split in two or three paragraphs.
AR: Introduction is corrected, and the first (and following) paragraph(s) are split.
RC: The authors should improve the Results section with the discussion of Tables 1 and 2 present in Section 4. And a more elaborate Discussion section should be provided.
AR: We improved the text, added discussion of the Tables and expand Discussion section according to the suggestions.
RC: The manuscript's novelty should be (more) explored in the Abstract, Introduction, and Conclusions.
AR: We have rewritten the text according to the suggestions.
We added the sentence in the Abstract to mark the novelty of the study:
“In contrast with classical OI, where only spatial correlations are considered, we developed the spatial-temporal optimal interpolation (STOI) technique for atmospheric applications with the use of spatial and temporal correlation functions.”
Author Response File: 
Author Response.pdf
Reviewer 2 Report
1: Minor English corrections, take note of past/present tense, is/are, phrase order, and prepositions. I suggest you use Grammarly and other proofing tools extensively.
Page 1
2: It seem from contributions that NM should be the first and corresponding author.
3: Abstract, generally ok but highlight AERONET is spatiotemporally sparse and the aim is to improve this by merging with GEOS-Chem predictions. Check past tense. the results (line 26-30) need to be rewritten to summarize your findings.
4: The introduction is missing references. Most sentences should be referenced here. Check past tense, active sentences (important bits first) e.g. line 38 Large differences are noted ....
Page 2
5: Data assimilation (lines 46 yo 53) is the crux of this paper provide more details
6: New paragraphs at line 54 Weighted coefficients... Why do you not consider time-based weighted coefficients, e.g. wildfires traffic peaks etc?
7: Split paragraph at line 57 KF is.. and again at 66 Variation methods....
8: The introduction needs to be further expanded. The techniques described need to be sufficiently detailed so that an educated reader unfamiliar with your topic can follow your logic.
Page 3
9: AERONET, what temporal frequency is this measured at? You mention Ve 3 but do not describe it.
10: Geo-Chem, describe your emission inventories in more detail. Is wildfire, shipping, aircraft, traffic, and biogenics included and at what temporal resolution? Why did you choose your horizontal grid resolution? It seems very coarse.
11: STOI description on page 4 is lacking critical information lines 145 to 153. Your methods should be sufficiently detailed so that the experiment can be repeated for a different region or simply to check your findings. The bottom of the page160-176 you calculate (lower case) k but this is not part of equations 1 & 2 and is not described.
Page 5
12: Results, lines 184-198 should be in methods. Why was the max length chosen as 850 km and 5 days?
13: Figure 3 needs a difference plot(c) and an explanation of the change. It appears to be predominantly changed to double the mid-AOD predicted values. How was the weighting function distributed in space
Page 6
14: This is the first time you mention the two months. I would suggest you need at a minimum all 12 months. I presume that the basis for these two months was a summer/winter difference, but this is not stated nor reached in the conclusions. Similarly lines 201 to 215 are mostly methods.
15: Figure 4 is a repeat of figure 1 and should be removed. Make the validation sites in Fig 1 large circles. (also consider that some readers may prefer a black and white printout [especially Fig 3] and test accordingly).
Page 7/8 Discussions and conclusions
16: This section needs a rewrite it is not substantiated by the methods. A typical reader may read the abstract then jump to this section to decide if they will read the entire article or if it's outside their field. What are the critical observations you made? [My summary of your findings are below]
a) there is more of an improvement over summer than winter and this is attributed to errors in the emission inventories potentially from the long-range transport of Saharan aerosols
b) There is more of an improvement at higher wavelengths, and this is attributed to...
c) the correlation plots showed that there was a concurrent improvement associated with denser AERONET sites Lille vs Minsk...
Line 265, correlation coefficient of what
Line 267-269 (besides the grammar) STOI is computationally efficient (where's the evidence). Significantly decreased RMSE (evidence?)
Author Response
Response to Reviewer 2 Comments
The revised manuscript atmosphere-2064524 by Miatselskaya et al. “Application of Optimal Interpolation to Spatially and Temporally Sparse Observations of Aerosol Optical Depth”
We thank Reviewer 2 for the valuable comments and suggestions that have helped to improve our manuscript. Our responses to the specific points are included below in blue.
Reviewer 2 Comments and Suggestions for Authors
Reviewer comment (RC).
1: Minor English corrections, take note of past/present tense, is/are, phrase order, and prepositions. I suggest you use Grammarly and other proofing tools extensively.
Authors Reply (AR). We have used Premium Grammarly proofing tools to improve English and to correct errors.
Page 1
2: It seem from contributions that NM should be the first and corresponding author.
AR: The contributions of the authors are provided in Author Contributions section.
3: Abstract, generally ok but highlight AERONET is spatiotemporally sparse and the aim is to improve this by merging with GEOS-Chem predictions. Check past tense. the results (line 26-30) need to be rewritten to summarize your findings.
AR: We have rewritten the abstract according to the suggestions:
“Aerosol optical depth (AOD) is one of the basic characteristics of atmospheric aerosol. A global ground-based network of sun and sky photometers, the Aerosol Robotic Network (AERONET), provides AOD data with low uncertainty. However, AERONET observations are sparse in space and time. To improve data density, we merged AERONET observations with a GEOS-Chem chemical transport model prediction using an optimal interpolation (OI) method. According to OI, we estimate AOD as a linear combination of observational data and a model forecast, with weight coefficients chosen to minimize a mean-square error in the calculation, assuming a negligible error of AERONET AOD observations. To obtain weight coefficients, we used correlations between model errors in different grid points. In contrast with classical OI, where only spatial correlations are considered, we developed the spatial-temporal optimal interpolation (STOI) technique for atmospheric applications with the use of spatial and temporal correlation functions. Using STOI, we obtained estimates of the daily mean AOD distribution over Europe. To validate the results, we compared daily mean AOD estimated by STOI with independent AERONET observations for two months and three sites. Compared with the GEOS-Chem model results, the averaged reduction of the root-mean-square error of the AOD estimate based on the STOI method is about 25%. The study shows that STOI provides a significant improvement in AOD estimates.”
4: The introduction is missing references. Most sentences should be referenced here. Check past tense, active sentences (important bits first) e.g. line 38 Large differences are noted ....
AR: The introduction has been updated with references and corrected passive voice to active voice where possible.
Page 2
5: Data assimilation (lines 46 yo 53) is the crux of this paper provide more details
AR: We have expanded the description of the data assimilation approach, and have provided the references to review papers on the subject.
6: New paragraphs at line 54 Weighted coefficients... Why do you not consider time-based weighted coefficients, e.g. wildfires traffic peaks etc?
AR: Weighting coefficients are calculated to obtain the best linear unbiased estimate. Data assimilation scheme takes into account the increase of AOD during wildfire without using corrections of weighting coefficients.
7: Split paragraph at line 57 KF is.. and again at 66 Variation methods....
AR: Done.
8: The introduction needs to be further expanded. The techniques described need to be sufficiently detailed so that an educated reader unfamiliar with your topic can follow your logic.
AR: The introduction has been expanded. However, we do not use Kalman filtering and variational methods in the present study, so we hope that it is no need to describe these techniques in details. We have added more references to allow readers to become more familiar with the mentioned methods.
Page 3
9: AERONET, what temporal frequency is this measured at? You mention Ve 3 but do not describe it.
AR: We have added the information on temporal frequency of AERONET measurements, the description of the Version 3, and the references.
10: Geo-Chem, describe your emission inventories in more detail. Is wildfire, shipping, aircraft, traffic, and biogenics included and at what temporal resolution? Why did you choose your horizontal grid resolution? It seems very coarse.
AR: We have added the description of GEOS-Chem emission inventories. We used the horizontal resolution of 0.25° latitude × 0.3125° longitude because this is the native resolution of the GEOS-FP meteorological fields. We have clarified this in the text. The used resolution is typical for chemical transport models.
11: STOI description on page 4 is lacking critical information lines 145 to 153. Your methods should be sufficiently detailed so that the experiment can be repeated for a different region or simply to check your findings. The bottom of the page160-176 you calculate (lower case) k but this is not part of equations 1 & 2 and is not described.
AR: We have added more details to the STOI description. We have added the definition of k.
Page 5
12: Results, lines 184-198 should be in methods. Why was the max length chosen as 850 km and 5 days?
AR: We re-arranged the text according to the suggestions and clarified the choice of the parameters in the Methods section.
13: Figure 3 needs a difference plot(c) and an explanation of the change. It appears to be predominantly changed to double the mid-AOD predicted values. How was the weighting function distributed in space
AR: Figure 3 is created for an illustration of the individual case of 17 July 2015. It intends to illustrate the result of the implementation of STOI. We have added some explanation of the AOD change. We do not use “weighting function”. The weighting coefficients are calculated using the correlation functions for every grid cell taking into account neighboring observations.
Page 6
14: This is the first time you mention the two months. I would suggest you need at a minimum all 12 months. I presume that the basis for these two months was a summer/winter difference, but this is not stated nor reached in the conclusions. Similarly lines 201 to 215 are mostly methods.
AR: We agree that we need more information to make better conclusions. We plan to obtain more results in future work. We have clarified this in the text. We have mentioned two months in the Abstract and in the Introduction.
15: Figure 4 is a repeat of figure 1 and should be removed. Make the validation sites in Fig 1 large circles. (also consider that some readers may prefer a black and white printout [especially Fig 3] and test accordingly).
AR: We believe that Figure 4 make it easier to follow the topic, shows clearly the neighboring AERONET sites and we prefer to leave it in the text.
Page 7/8 Discussions and conclusions
16: This section needs a rewrite it is not substantiated by the methods. A typical reader may read the abstract then jump to this section to decide if they will read the entire article or if it's outside their field. What are the critical observations you made? [My summary of your findings are below]
- a) there is more of an improvement over summer than winter and this is attributed to errors in the emission inventories potentially from the long-range transport of Saharan aerosols
- b) There is more of an improvement at higher wavelengths, and this is attributed to...
- c) the correlation plots showed that there was a concurrent improvement associated with denser AERONET sites Lille vs Minsk...
AR: We have rewritten the section taking into account most of suggestions.
Line 265, correlation coefficient of what
AR: We explained this more clearly:
“In [43], we evaluated the correlation between the observed AOD and the modeled by GEOS-Chem AOD values, using statistics of observations at the 88 European AERONET sites and for the 2015–2016 period. The correlation coefficients obtained in [43] were 0.60, 0.53, and 0.45 for wavelengths 440, 675, and 870 nm, respectively.”
Line 267-269 (besides the grammar) STOI is computationally efficient (where's the evidence). Significantly decreased RMSE (evidence?)
AR: We have rewritten the text according to the suggestions. The results show a significant decrease in the RMSE when STOI is applied. The grammar is checked and the text improved.
Author Response File: Author Response.pdf
Reviewer 3 Report
The paper proposes the spatial-temporal optimal interpolation method for evaluation of atmospheric aerosol load in locations where the direct measurements on site are unavailable.
The results obtained are interesting but before publication the following improvements must be performed:
1) In the Introduction the reader must be informed for several applications made in the past investigating the changes in AOD, e.g.,. For some of these earthquake events, abnormal increases in time series of AOD data were detected before and after the occurrence of the earthquake. Nevertheless, in other cases, no clear anomalies were observed around the earthquake date. See: I "On the association of aerosol optical depth and total ozone fluctuations with recent earthquakes in Greece." Acta Geophysica 65.4 (2017): 659-665.
2) Related to the previous comment is the earlier investigation made for the soiling estimations of historic and modern materials by making use of PM10 concentration calculated using satellite AOD,. See: "The deterioration of materials as a result of air pollution as derived from satellite and ground based observations." Atmospheric Environment 185 (2018): 91-99.
"The enhanced deterioration of the cultural heritage monuments due to air pollution." Environmental Science and Pollution Research 16.5 (2009): 590-592.
"An observational study of the atmospheric ultra-fine particle dynamics." Atmospheric Environment 59 (2012): 312-319.
3) To attract the reader's interest, add a very brief discussion on what is the contribution of your research to the success of the Sustainable Development Goals-UN 2030 agenda. See: "Remote Sensing Letters contribution to the success of the Sustainable Development Goals-UN 2030 agenda." Remote Sensing Letters 11.8 (2020): 715-719.
In conclusion, I would be happy to review the revised version of the manuscript, believing that the above-mentioned suggestions will substantially improve it and then it will be worthy of publication.
Author Response
Response to Reviewer 3 Comments
The revised manuscript atmosphere-2064524 by Miatselskaya et al. Application of Optimal Interpolation to Spatially and Temporally Sparse Observations of Aerosol Optical Depth
We thank Reviewer 3 for the valuable comments and suggestions that have helped to improve our manuscript. Our responses to the specific points are included below in blue.
Reviewer 3 Comments and Suggestions for Authors
Moderate English changes required
Authors Reply (AR). We improved our English where possible and used Premium Grammarly as well.
Reviewer comment (RC).
The paper proposes the spatial-temporal optimal interpolation method for evaluation of atmospheric aerosol load in locations where the direct measurements on site are unavailable.
The results obtained are interesting but before publication the following improvements must be performed:
RC: 1) In the Introduction the reader must be informed for several applications made in the past investigating the changes in AOD, e.g.,. For some of these earthquake events, abnormal increases in time series of AOD data were detected before and after the occurrence of the earthquake. Nevertheless, in other cases, no clear anomalies were observed around the earthquake date. See: I "On the association of aerosol optical depth and total ozone fluctuations with recent earthquakes in Greece."Acta Geophysica 65.4 (2017): 659-665.
AR: The text is updated, and the suggested reference is included in the Introduction.
RC: 2) Related to the previous comment is the earlier investigation made for the soiling estimations of historic and modern materials by making use of PM10 concentration calculated using satellite AOD,. See: "The deterioration of materials as a result of air pollution as derived from satellite and ground based observations."Atmospheric Environment 185 (2018): 91-99.
AR: The text is updated, and the suggested reference is included.
RC: "The enhanced deterioration of the cultural heritage monuments due to air pollution."Environmental Science and Pollution Research 16.5 (2009): 590-592.
AR: The text is updated, and the suggested reference is included.
RC: "An observational study of the atmospheric ultra-fine particle dynamics." Atmospheric Environment 59 (2012): 312-319.
AR: It is difficult to include the proposed reference in the text because it is far from the topic of the manuscript.
RC: 3) To attract the reader's interest, add a very brief discussion on what is the contribution of your research to the success of the Sustainable Development Goals-UN 2030 agenda. See: "Remote Sensing Letters contribution to the success of the Sustainable Development Goals-UN 2030 agenda." Remote Sensing Letters 11.8 (2020): 715-719.
AR: That is a good idea in general. However, the topic of the manuscript is focused on methods of data assimilation and is not connected directly with the Sustainable Development Goals-UN 2030 agenda.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The authors responded satisfactorily to all my suggestions and considerably improved their manuscript. I recommend it to be accepted in its present form.
Reviewer 2 Report
The authors have revised the article and have addressed the issues raised.
Reviewer 3 Report
I went through the revised version and found it much improved.
Therefore, I recommend its publication as is.
