the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 21 Dec 2022
Evaluation of total ozone measurements from Geostationary Environmental Monitoring Satellite (GEMS)
Kanghyun Baek
Juseon Bak
David P. Haffner
Mina Kang
Hyunkee Hong
Abstract. As all life on earth depends crucially on atmospheric ozone, low earth orbiting (LEO) satellites have been used to monitor atmospheric ozone to reduce its impact on the environment and public health. The continued interest in air pollution and stratospheric ozone variability has motivated the development of a geostationary environmental monitoring satellite (GEMS) for hourly ozone monitoring. This paper provides the atmospheric science community with the world's first assessment of GEMS total column ozone (TCO) retrieval performance and diurnal ozone variation. The algorithm used for GEMS is a more advanced version of its predecessor, the TOMS-V8 algorithm. In addition to calculating total ozone, the algorithm has the advantage of providing ozone profile and retrieval error information. To assess the performance of the GEMS algorithm, the hourly GEMS total ozone was compared with ground-based measurements from four Pandora instruments and other satellite platforms from TROPOMI and OMPS. A high correlation of 0.91 or more with GEMS and Pandora TCO at Seoul, Busan, and Yokosuka but a low correlation of 0.83 at Ulsan, which is significantly smaller than at other sites. Root-mean-squared error (RMSE) showed satisfactory small values, with the lowest RMSE of 2.06 DU. Positive mean biases (MBs) were observed at all sites. This agreement suggests that the GEMS hourly ozone monitoring allows for continuous updates about stratospheric ozone and its related atmospheric changes. The quantitative comparison of GEMS TCO data with TROPOMI and OMPS TCO data shows a high correlation coefficient greater than 0.98 and a low RMSE of less than 1.8 DU over clear sky conditions. GEMS TCO underestimates by - 0.14 % (0.4 DU) with a standard deviation of 2.0 % relative to TROPOMI and overestimates by + 0.1 % (0.3 DU) with a standard deviation of 2.3 % relative to OMPS. It shows that the GEMS TCO agrees very well with the TROPOMI and OMPS TCO. The results are a meaningful scientific advance by providing the first validated, hourly UV ozone retrievals from a satellite in geostationary orbit. This experience can be used to advance research with future geostationary environmental satellite missions, including incoming TEMPO and Sentinel-4.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(1214 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1214 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Kanghyun Baek et al.
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-1402', Anonymous Referee #1, 27 Jan 2023
The manuscript present total ozone results from the GEMS geostationary satellite instrument, and compares them with independent measurements from other satellite as well as from ground based remote sensing. Comparisons with ground-based data are generally good, however, it reveals a time-dependent drift, which may be instrumental and/or seasonal. Comparisons with OMPS and TROPOMI showed high correlation and low bias, however a latitude dependent error in the GEMS data.
The paper gives a good overview of the current state of the GEMS total ozone data, however the algorithm description needs significant expansion and additional analysis is recommended to provide a better insight in the reported drifts between GEMS and the validation data.
Major points
The algorithm that has been developed for GEMS has not been published in the open literature. The algorithm description in the current paper leaves many aspects unanswered. Although it is based on a well-known total ozone algorithm, specific aspects import for a GEO instrument versus a LEO instrument are not addressed. I recommend significantly expanding section 2.2 to include the following aspects:
- The algorithm uses a LUT based radiative transfer forward model. Provide an assessment of the error that this LUT based forward model makes wrt an online RTM (VLIDORT) and how this error propagates to total ozone.
- Provide in a supplemental section a full description of the LUT RTM, its dimensions and methods used for interpolating this LUT. Also, I think this LUT and tools to interpolate it should be made available.
- A unique aspect of GEMS is the hourly observations. However, geometries vary strongly over the GEMS field-of-view. What is the expected effect of the viewing geometries on the vertical sensitivity of the ozone observations. How does the averaging kernel vary of the FOV and over time of the day? This is important information to understand the GEMS observations and the difference with LEO observations.
- What is the impact of the choice of a-priori ozone profiles and the assumed a-priori errors? This especially important as you are fitting an ozone profile with 11 layers, using only 3 wavelengths. Hence the retrieval is heavily underdetermined and thus depending on a-prior information.
The abstract leaves out important findings of the validation. Specifically, the time dependent drift and the latitudinal dependent errors shall be mentioned in the abstract.
Minor points
In figure 4 comparisons are shown for GEMS, TROPOMI and OMPS. I propose to include in the figure (or in a supplemental figure) the results of the GEMS-Pandora comparison at the mean overpass time of TROPOMI/OMPS. In this way potential errors that very over the day are not folded into this comparison, and the comparison with TROPOMI and OMPS is much cleaner.
What is the status of the GEMS data set? Is produced by the operational processor and available for users?
For all datasets (GEMS, OMPS, TROPOMI, Pandora), the version used in the work should be clearly documented. When available the doi of the dataset should be used.
Figure 11 and 12 appear exactly the same to me. Is by mistake the wrong figure used in the manuscript?
To overcome issues with the calibration of the solar spectrum, I would suggest processing (part of) the GEMS data with a fixed solar spectrum. What is the impact on the seasonality if this approach is followed?
In the conclusions the authors mention that the ozone data is expected to improve by improving the GEMS characterization. What is the timeline for this. How is this coupled to public data release and/or version of the GEMS data?
Citation: https://doi.org/10.5194/egusphere-2022-1402-RC1 -
AC1: 'Reply on RC1', Jae H. Kim, 14 Apr 2023
Thank you very much for taking the time to review our manuscript. During the review process, we have replaced the results analyzed from GEMS V1.0 with GEMS V2.0 data. We found that there were some errors in the LUT calculations used for ozone calculations in V1.0, which could affect the accuracy of the results. Therefore, we replaced all analysis data with V2.0 to prevent any such errors. As a result, GEMS V2.0 shows about a 2% lower ozone calculation result compared to V1.0, and all verification metrics of the analysis results have changed.
We appreciate your valuable comments and suggestions, and we have addressed each of your concerns in the revised version of manuscript and supplementary material. Please find our detailed response below.
-
RC2: 'Comment on egusphere-2022-1402', Anonymous Referee #2, 04 Feb 2023
Review of Baek et al.
This manuscript describes validation of the world’s first geostationary-satellite-based total ozone observations over Asia, achieved by GEMS, using Pandora spectrometers, and comparisons with TROPOMI/OMPS satellite observations. Basically its high performance was confirmed, with high correlation coefficients and small biases. I find the methodology and logics are almost sound to draw this main conclusion. However, more explanation and justification are needed to support details. First, though the authors state that retrieval of diurnal variation (line 15) and providing retrieval error information (line 17) are the features with GEMS, validation and comparisons are not given as a function of time of day and no discussion about the error of retrievals is made. Second, only overall "positive" mean bias is mentioned in the abstract but in detail the bias is rather strong negative (up to -6%) for the mid/high latitudes. This is mentioned in conclusion but should be mentioned in Abstract as well. Third, algorithm versions or product names from TROPOMI, OMPS, and Pandora are lacking and thus the results are not traceable. To consider publication, major revisions are required about the points above and the specific points listed below.
- Line 16. Be clear in which aspect the GEMS retrieval is advanced. Maybe those listed in lines 60-61. Mention them in short here.
- Lines 15 and 17. Results of the retrieval error information should be discussed in the main text. Biases should be analyzed and depicted as a function of time of day, as the diurnal observation capability is highlighted.
- Lines 21 and 27. Small positive biases and "very well agreement" are mentioned but in reality negative biases for mid/latitudes are found against satellites and Pandora. This should be described with a good balance.
- Line 115, 21 ozone profiles are mentioned but how this is applied is not very clear, particularly with the statement of "three to ten ozone profiles" in line 152.
- Section 2.2.2 and Figure 1. Step 1, 2, and 3 should be mentioned in Figure 1 caption. Maybe red, green and blue parts are the steps, individually.
- Line 216. What is the "situation"?
- Line 223. TROPOMI
- Section 3.2. Need to mention algorithm versions or product names for Pandora, TROPOMI, and OMPS. Acknowledgments to the PIs need to be included.
- Table 1. Slash characters are required to separate month and day at several positions.
- Line 251. Remove "However,"
- Line 255. This decrease (likely the one shown in Figure 5 and 11) could be seasonal (as mentioned in conclusion) or long-term degrading trend (as implied here)?
- Line 271. The statement that Pandora uses a fixed-temperature ozone absorption coefficient needs to be checked. In the recent algorithm version 1.8, the products "out2" for example considers the temperature dependence as climatology. For this perspective, mentioning algorithm version/product name is necessary for the traceability.
- Line 297. Rewrite the sentence starting with "These bad pixels ..."
- Line 332. Are the -0.14 +/- 2.00 % and +0.10+/-2.31% mean biases?
- Line 344. Perhaps Nishinoshima?
- Line 364. Are the distinct spatial and seasonal variability relevant to the features of the bias discussed here?
- Figure 12. No difference is found with Figure 11.
Citation: https://doi.org/10.5194/egusphere-2022-1402-RC2 -
AC2: 'Reply on RC2', Jae H. Kim, 14 Apr 2023
Thank you very much for taking the time to review our manuscript. During the review process, we have replaced the results analyzed from GEMS V1.0 with GEMS V2.0 data. We found that there were some errors in the LUT calculations used for ozone calculations in V1.0, which could affect the accuracy of the results. Therefore, we replaced all analysis data with V2.0 to prevent any such errors. As a result, GEMS V2.0 shows about a 2% lower ozone calculation result compared to V1.0, and all verification metrics of the analysis results have changed.
We appreciate your valuable comments and suggestions, and we have addressed each of your concerns in the revised version of manuscript and supplementary material. Please find our detailed response below.
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-1402', Anonymous Referee #1, 27 Jan 2023
The manuscript present total ozone results from the GEMS geostationary satellite instrument, and compares them with independent measurements from other satellite as well as from ground based remote sensing. Comparisons with ground-based data are generally good, however, it reveals a time-dependent drift, which may be instrumental and/or seasonal. Comparisons with OMPS and TROPOMI showed high correlation and low bias, however a latitude dependent error in the GEMS data.
The paper gives a good overview of the current state of the GEMS total ozone data, however the algorithm description needs significant expansion and additional analysis is recommended to provide a better insight in the reported drifts between GEMS and the validation data.
Major points
The algorithm that has been developed for GEMS has not been published in the open literature. The algorithm description in the current paper leaves many aspects unanswered. Although it is based on a well-known total ozone algorithm, specific aspects import for a GEO instrument versus a LEO instrument are not addressed. I recommend significantly expanding section 2.2 to include the following aspects:
- The algorithm uses a LUT based radiative transfer forward model. Provide an assessment of the error that this LUT based forward model makes wrt an online RTM (VLIDORT) and how this error propagates to total ozone.
- Provide in a supplemental section a full description of the LUT RTM, its dimensions and methods used for interpolating this LUT. Also, I think this LUT and tools to interpolate it should be made available.
- A unique aspect of GEMS is the hourly observations. However, geometries vary strongly over the GEMS field-of-view. What is the expected effect of the viewing geometries on the vertical sensitivity of the ozone observations. How does the averaging kernel vary of the FOV and over time of the day? This is important information to understand the GEMS observations and the difference with LEO observations.
- What is the impact of the choice of a-priori ozone profiles and the assumed a-priori errors? This especially important as you are fitting an ozone profile with 11 layers, using only 3 wavelengths. Hence the retrieval is heavily underdetermined and thus depending on a-prior information.
The abstract leaves out important findings of the validation. Specifically, the time dependent drift and the latitudinal dependent errors shall be mentioned in the abstract.
Minor points
In figure 4 comparisons are shown for GEMS, TROPOMI and OMPS. I propose to include in the figure (or in a supplemental figure) the results of the GEMS-Pandora comparison at the mean overpass time of TROPOMI/OMPS. In this way potential errors that very over the day are not folded into this comparison, and the comparison with TROPOMI and OMPS is much cleaner.
What is the status of the GEMS data set? Is produced by the operational processor and available for users?
For all datasets (GEMS, OMPS, TROPOMI, Pandora), the version used in the work should be clearly documented. When available the doi of the dataset should be used.
Figure 11 and 12 appear exactly the same to me. Is by mistake the wrong figure used in the manuscript?
To overcome issues with the calibration of the solar spectrum, I would suggest processing (part of) the GEMS data with a fixed solar spectrum. What is the impact on the seasonality if this approach is followed?
In the conclusions the authors mention that the ozone data is expected to improve by improving the GEMS characterization. What is the timeline for this. How is this coupled to public data release and/or version of the GEMS data?
Citation: https://doi.org/10.5194/egusphere-2022-1402-RC1 -
AC1: 'Reply on RC1', Jae H. Kim, 14 Apr 2023
Thank you very much for taking the time to review our manuscript. During the review process, we have replaced the results analyzed from GEMS V1.0 with GEMS V2.0 data. We found that there were some errors in the LUT calculations used for ozone calculations in V1.0, which could affect the accuracy of the results. Therefore, we replaced all analysis data with V2.0 to prevent any such errors. As a result, GEMS V2.0 shows about a 2% lower ozone calculation result compared to V1.0, and all verification metrics of the analysis results have changed.
We appreciate your valuable comments and suggestions, and we have addressed each of your concerns in the revised version of manuscript and supplementary material. Please find our detailed response below.
-
RC2: 'Comment on egusphere-2022-1402', Anonymous Referee #2, 04 Feb 2023
Review of Baek et al.
This manuscript describes validation of the world’s first geostationary-satellite-based total ozone observations over Asia, achieved by GEMS, using Pandora spectrometers, and comparisons with TROPOMI/OMPS satellite observations. Basically its high performance was confirmed, with high correlation coefficients and small biases. I find the methodology and logics are almost sound to draw this main conclusion. However, more explanation and justification are needed to support details. First, though the authors state that retrieval of diurnal variation (line 15) and providing retrieval error information (line 17) are the features with GEMS, validation and comparisons are not given as a function of time of day and no discussion about the error of retrievals is made. Second, only overall "positive" mean bias is mentioned in the abstract but in detail the bias is rather strong negative (up to -6%) for the mid/high latitudes. This is mentioned in conclusion but should be mentioned in Abstract as well. Third, algorithm versions or product names from TROPOMI, OMPS, and Pandora are lacking and thus the results are not traceable. To consider publication, major revisions are required about the points above and the specific points listed below.
- Line 16. Be clear in which aspect the GEMS retrieval is advanced. Maybe those listed in lines 60-61. Mention them in short here.
- Lines 15 and 17. Results of the retrieval error information should be discussed in the main text. Biases should be analyzed and depicted as a function of time of day, as the diurnal observation capability is highlighted.
- Lines 21 and 27. Small positive biases and "very well agreement" are mentioned but in reality negative biases for mid/latitudes are found against satellites and Pandora. This should be described with a good balance.
- Line 115, 21 ozone profiles are mentioned but how this is applied is not very clear, particularly with the statement of "three to ten ozone profiles" in line 152.
- Section 2.2.2 and Figure 1. Step 1, 2, and 3 should be mentioned in Figure 1 caption. Maybe red, green and blue parts are the steps, individually.
- Line 216. What is the "situation"?
- Line 223. TROPOMI
- Section 3.2. Need to mention algorithm versions or product names for Pandora, TROPOMI, and OMPS. Acknowledgments to the PIs need to be included.
- Table 1. Slash characters are required to separate month and day at several positions.
- Line 251. Remove "However,"
- Line 255. This decrease (likely the one shown in Figure 5 and 11) could be seasonal (as mentioned in conclusion) or long-term degrading trend (as implied here)?
- Line 271. The statement that Pandora uses a fixed-temperature ozone absorption coefficient needs to be checked. In the recent algorithm version 1.8, the products "out2" for example considers the temperature dependence as climatology. For this perspective, mentioning algorithm version/product name is necessary for the traceability.
- Line 297. Rewrite the sentence starting with "These bad pixels ..."
- Line 332. Are the -0.14 +/- 2.00 % and +0.10+/-2.31% mean biases?
- Line 344. Perhaps Nishinoshima?
- Line 364. Are the distinct spatial and seasonal variability relevant to the features of the bias discussed here?
- Figure 12. No difference is found with Figure 11.
Citation: https://doi.org/10.5194/egusphere-2022-1402-RC2 -
AC2: 'Reply on RC2', Jae H. Kim, 14 Apr 2023
Thank you very much for taking the time to review our manuscript. During the review process, we have replaced the results analyzed from GEMS V1.0 with GEMS V2.0 data. We found that there were some errors in the LUT calculations used for ozone calculations in V1.0, which could affect the accuracy of the results. Therefore, we replaced all analysis data with V2.0 to prevent any such errors. As a result, GEMS V2.0 shows about a 2% lower ozone calculation result compared to V1.0, and all verification metrics of the analysis results have changed.
We appreciate your valuable comments and suggestions, and we have addressed each of your concerns in the revised version of manuscript and supplementary material. Please find our detailed response below.
Peer review completion
Journal article(s) based on this preprint
Kanghyun Baek et al.
Kanghyun Baek et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 492 | 282 | 29 | 803 | 16 | 19 |
- HTML: 492
- PDF: 282
- XML: 29
- Total: 803
- BibTeX: 16
- EndNote: 19
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1214 KB) - Metadata XML