the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Jul 2022
Carbon dioxide cover: carbon dioxide column concentration seamlessly distributed globally during 2009–2020
Abstract. For carbon dioxide concentration (XCO2) distribution, the improvement of spatial and temporal resolution is very important in some scientific studies (e.g., studies of the carbon cycle and assessment of carbon emissions based on top-down theory). However, carbon sniffing satellites based on passive theory (e.g., Gosat-2, OCO-2, and OCO-3) are susceptible to cloud and aerosol interference when the data are captured. Therefore, the data collected by carbon sniffing satellites have relatively low utilization, especially in some regions where data gaps exist. Here, we present the Carbon Dioxide Coverage (CDC) dataset, an innovative theory to obtain high spatial and temporal resolution maps of XCO2 distribution by combining spatial attributes and extracted temporal attributes from the GOSAT satellite series data. This theory is divided into the following three parts. Firstly, several background values in the raw GOSAT data were removed through data pre-processing, and for spatial attributes, GOSAT satellite data gap areas were filled by combining adjacent GOSAT data and empirical Bayesian kriging (EBK) theory in the study area. Secondly, for the temporal attributes, we constructed a time profile parameter library, based on the GOSAT data of the time series to extract the temporal parameters from a specific formula at each point of the study area. Finally, for the integration of temporal and spatial information, based on the GOSAT satellite data and the populated data based on spatial attributes, we assign the temporal parameter information from the time parameter library to each pixel location in the study area, combining the transfer component analysis (TCA) theory, and then combine the assigned parameters with specific formulas to complete the prediction of XCO2 distribution. For temporal resolution, both the GOSAT_FTS_L3_V2.95 and CDC datasets are monthly-averaged resolution datasets from 2010 to 2020. And for spatial resolution, the CDC dataset is 0.25° resolution with a significant improvement compared to GOSAT_FTS_L3_V2.95 which is 2.5° resolution. And the dataset contained 136 files. Besides, for the data validation part, we used OCO-2 satellite data from 2009 to 2020 and TCCON data at mid and low latitudes, respectively. This CDC dataset and the original data from the TCCON sites were compared on a monthly-averaged scale. And the results showed that R2 was 0.9686, and RMSE was 1.3811 ppm. We also derived statistical monthly averaged XCO2 from OCO-2 data and compared it with the data set from our theory. And our evaluation index R was greater than 0.7, by comparison with OCO-2 during 2014–2020. Finally, to assess the accuracy of the algorithm, we compared the predicted results with the input data for the period of 2009–2020. And the comparison results show that the mean value of R2 is 0.93 and the mean value of RMSE is 0.53 ppm during 2010–2020. Data gaps produced by sniffer satellites are disturbed by factors such as clouds and aerosols and can be filled by this mapping technique is mentioned in this paper. This technique improves the utilization of XCO2 and the accuracy and resolution of the CDC dataset is sufficient for scientific applications. And the CDC dataset is publicly available at https://doi.org/10.6084/m9.figshare.17826404.v4 (Zhang et al., 2022), which is of significance for a multitude of scientific carbon research.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(8352 KB)
Interactive discussion
Status: closed
-
RC1: 'Comment on essd-2022-215', Anonymous Referee #1, 13 Oct 2022
<General Comments>
Long-term and regional scale monitoring of CO2 from space is important for understanding climate changes. Satellite can cover globally but clear sky ratios vary much region by region. Spatial-temporal technique are useful. However, I found several critical issues in this paper. I recommend resubmission.
(1) GOSAT sampling pattern and CO2 density enhancement over large emission sources
The GOSAT sampling pattern consists of grid observation and target observation. The sampling pattern is not uniform. GOSAT is targeting global megacities which shows local enhancement. Over the ocean, GOSAT is tracking the specular reflection points of the sun, of which sampling pattern is not grid. Authors should describe in more detail how to use these data for analysis.
(2) GOSAT data source
I do not understand why authors use the NIES level 3 products.
Level 3 products are spatially interpolated already. As mentioned in (1), they are problematic. There are several Level 2 GOSAT products other than NIES such as ACOS, RemoTeC, University of Leicester, and JAXA. Why do authors use NIES products? There is no product defined as “official”.
<Specific Comments>
(1) 2.2 validation data TCCON
When the authors use the muti-year data in TCCON comparison, the coefficient of determination becomes too good. The annual growth of global CO2 density should be removed for the analysis. The deviation and bias of matched up data should be presented.
(2) 2.2 Validation data: OCO-2 Level 2 product
The version of the OCO-2 level 2 products should be described. Older OCO-2 products have topography dependent bias. The difference in footprints of GOSAT and OCO-2 creates errors.
<Technical Corrections>
(1) Page 15 Table 1
Is the unit of RMSE ppm?
(2) page 19, Figure 2
The branching of left and right should be described.
Citation: https://doi.org/10.5194/essd-2022-215-RC1 - AC1: 'Reply on RC1', Xin Ma, 09 Nov 2022
- AC3: 'Reply on RC1', Xin Ma, 27 Nov 2022
-
CC1: 'Comment on essd-2022-215', Mengqi Zhang, 09 Nov 2022
In this paper, a spatiotemporal interpolation method is developed, and a data product with full spatiotemporal coverage is generated by using the XCO2 data of GOSAT.I'm not particularly aware of the article category for ESSD, but compared to other research papers in ESSD, it's more suitable for technical description articles, at least at this stage. The article has some obvious scientific errors and inappropriate knowledge descriptions, the following points should be considered to improve the quality of the article, especially some major errors. As community comments, I believe these comments will increase the understanding of carbon monitoring satellite data assimilation and improve this research.
Although we are very concerned about the carbon cycle and the spatiotemporal distribution of CO2, for atmospheric inversion models, sparse data observations are sufficient to obtain carbon fluxes. NOTE I'm not denying that we don't need a spatially seamlessly CO2 distribution, but the introduction should explain why we need a spatiotemporally seamlessly CO2, such as calculating global averages, analyzing seasonal changes.
P6. Line 160. Equation 1 looks very strange and seems wrong. What is the difference between coefficients c and e? What is the difference between d and g? The result needs to be checked carefully and even recalculated. In addition, adopting this method to construct time series would lead to significant drawbacks. I believe the authors may not understand the CO2 growth rate, please see papers such as Buchwitz. et.al, 2018 and A. Chatterjee et.al, 2017 to understand the significance of CGR in reflecting vegetation, climate, etc. The method of presetting a function to fit will not be able to capture the real change in the CO2 growth rate, and the function will be directly known after derivation. One of the reasons we want seamless data is to better calculate the global average, and thus the growth rate, that is, the net flux. Fitting with a fixed function would loss this.
In the abstract and the introduction, the authors claim that the amount of XCO2 data is mainly affected by factors such as clouds and aerosols. This is wrong, it's actually a swath issue. The design of the width is related to the fluctuation amplitude of atmospheric CO2 and optical inversion factors. Authors are advised to read the relevant literature and correct the description in this section. While this is not that important for this article, the readers needs to understand the real background for this work.
P3, L65-83. This section overlaps with method descriptions in subsequent part, and it is not appropriate to introduce too much about the methods of this study in the introduction.
P4, L98 "...the accuracy of the comparison between the GOSAT data product and the TCCON site was 0.56 ppm", it is not appropriate to use the "accuracy" word, it should be stated, such as standard deviation, bias, etc.
P4, L100, "... three-day temporal resolution. The time resolution of GOSAT-2 satellite is 6 days..." is inappropriate. The correct description is the revisit cycle/repeat cycle. Please differentiate these concepts.(temporal resolution,time resolution, repeat cycle,)
P4, L105. However, the OCO-2_L2_Lite_FP9r provides data locations that are gradually shifted over time by satellite observations. This sentence is difficult to understand. I think you should express that the orbits of the sub-satellite points are evenly distributed? Illustration may be needed.
P4, L112 column-averaged XCO2. This is wrong. And the full name of XCO2 is wrong, including the title, abstract and etc. I think it shold be carefully checked the full text of the corresponding full scientific name.
P4. L107, "fixed location" should be more clear. L109. the six data channels are wrong. TANSO-FTS is a 4–band interferometer.
P4., L123. Please correct for column-averaged abundances of CO2 expression. And the results showed that R2 was 0.9686, and RMSE was 1.3811 . Please indicate the source.
P5. 146 EBK theory or EBK method ? Please express it in a unified way. such as L150.
In Section 2.4, more indicators for accuracy evaluation should be added. Bias and standard deviation are necessary in the verification of XCO2.
P7. 195. It is not necessary to show the results of each year, only a few examples, such as some months or a specific year, are sufficient. From the analysis of the data, such as some seasonal changes, changes in CO2 growth, and spatial differences may be more meaningful.
Figure 6, As said at the beginning, of course we know that CO2 is rising, but its growth rate is more meaningful, and it is recommended to draw a related graph of the growth rate. If it does not reflect reasonable fluctuations, but a fully sinusoidal pattern, the study would be significantly flawed.
From the research point of view, averaging kernel and the prior profile should be considered in comparison with OCO-2. Although they may be ignored in some cases and not important on monthly validation where accuracy is not required, the article should mention it.
Figure 3,4 and 5. PXCO2 TXCO2 P XCO2 should be described uniformly (note the space). Other than that, I would suggest that it would be better to do time series validation on a monthly basis. Specifically, the horizontal axis is time, and the vertical axis is parameters such as error, which can also be filled with error distribution, which is more intuitive.
Figure 7~17. It looks like this resolution may be trapped in a highly smooth phenomenon, which means it may not really be 0.25 degrees. It is recommended to draw a detailed map of some regions to show that the method does have this good resolution and can capture reasonable and sufficient spatial gradient changes. It is also recommended to compare the results of models, such as CarbonTracker or the L4B model products of GOSAT-NIES, to demonstrate the rationality of the results.
As mentioned above, in addition to the error evaluation in the time dimension, the error in the spatial dimension should also be evaluated to illustrate the reliability of the data.
I am not an Native English speaker. But I believe that the English of this article should be greatly improved, including scientific names of many nouns, and descriptions in scientific language.
Buchwitz, Michael, et al. "Computation and analysis of atmospheric carbon dioxide annual mean growth rates from satellite observations during 2003–2016." Atmospheric Chemistry and Physics 18.23 (2018): 17355-17370.
Chatterjee, A., et al. "Influence of El Niño on atmospheric CO2 over the tropical Pacific Ocean: Findings from NASA’s OCO-2 mission." Science 358.6360 (2017): eaam5776.
Citation: https://doi.org/10.5194/essd-2022-215-CC1 -
AC2: 'Reply on CC1', Xin Ma, 27 Nov 2022
Dear reviewer, thank you for your kindly suggestions. For specific responses, please refer to the attachment.
-
CC2: 'Reply on AC2', Mengqi Zhang, 27 Nov 2022
The authors put a lot of effort into addressing most of my concerns and suggestions. The authors address most of the errors and concerns, especially fixing Equation 1. The author added more validation and analysis content, which greatly improved the quality of the article.
As a community comment, overall, the quality of the article has been greatly improved. I think it meets the basic criteria for publishing on ESSD at least for now.
But some content still needs polishing. I have some suggestions as follows.
On the response of "P4, L105. However, the OCO-2_L2_Lite_FP9r provides data locations that are gradually shifted over
time by satellite observations", I already understand what the author wants to convey. But in fact, "the satellite orbit is gradually offset in order
to collect more data." is the Inherent Properties of Sun-Synchronous Orbits(SSO). Most polar-orbiting remote sensing satellites, such as Terra and Aqua, opt for SSO orbits, which allow them to image the globe in a single day. For OCO-2, it is also an SSO orbit, but its width is so narrow that it can only detect sub-satellite points (there are also SAM mode and glint mode). Here I just want to remind, not deny anything.If "Fig. 13-1 The distribution of seasonal mean XCO2 in 2010" is added to the revised manuscript, it is better to replace abcd with the name of the season, or the abbreviation of the month (eg JFM).
If "Figure 16-1 Monthly-averaged XCO2 validation results for TCCON data and CDC dataset at global mid- and low-latitude TCCON sites from 200906 to 202012. " is added to the revised manuscript, it is recommended to check the color fill of the standard deviation. Although readers can understand its meaning, 1x std and 2x std seem to be reversed, maybe it may be a color overlay problem, it is recommended to check, including others.
According to ESSD's publication policy (https://essd.copernicus.org/articles/10/2275/2018/, ), Acknowledgements should mention the OCO-2 and GOSAT data used in this article, as well as the TCCON data, and other publicly available data used. This is also a general requirement.
Overall, the author has solved most of the problems, and I believe the above comments can better improve the quality of the article. As a community comment, I have no further comments.
Citation: https://doi.org/10.5194/essd-2022-215-CC2
-
CC2: 'Reply on AC2', Mengqi Zhang, 27 Nov 2022
-
AC2: 'Reply on CC1', Xin Ma, 27 Nov 2022
- RC2: 'Comment on essd-2022-215', Anonymous Referee #2, 26 Nov 2022
Interactive discussion
Status: closed
-
RC1: 'Comment on essd-2022-215', Anonymous Referee #1, 13 Oct 2022
<General Comments>
Long-term and regional scale monitoring of CO2 from space is important for understanding climate changes. Satellite can cover globally but clear sky ratios vary much region by region. Spatial-temporal technique are useful. However, I found several critical issues in this paper. I recommend resubmission.
(1) GOSAT sampling pattern and CO2 density enhancement over large emission sources
The GOSAT sampling pattern consists of grid observation and target observation. The sampling pattern is not uniform. GOSAT is targeting global megacities which shows local enhancement. Over the ocean, GOSAT is tracking the specular reflection points of the sun, of which sampling pattern is not grid. Authors should describe in more detail how to use these data for analysis.
(2) GOSAT data source
I do not understand why authors use the NIES level 3 products.
Level 3 products are spatially interpolated already. As mentioned in (1), they are problematic. There are several Level 2 GOSAT products other than NIES such as ACOS, RemoTeC, University of Leicester, and JAXA. Why do authors use NIES products? There is no product defined as “official”.
<Specific Comments>
(1) 2.2 validation data TCCON
When the authors use the muti-year data in TCCON comparison, the coefficient of determination becomes too good. The annual growth of global CO2 density should be removed for the analysis. The deviation and bias of matched up data should be presented.
(2) 2.2 Validation data: OCO-2 Level 2 product
The version of the OCO-2 level 2 products should be described. Older OCO-2 products have topography dependent bias. The difference in footprints of GOSAT and OCO-2 creates errors.
<Technical Corrections>
(1) Page 15 Table 1
Is the unit of RMSE ppm?
(2) page 19, Figure 2
The branching of left and right should be described.
Citation: https://doi.org/10.5194/essd-2022-215-RC1 - AC1: 'Reply on RC1', Xin Ma, 09 Nov 2022
- AC3: 'Reply on RC1', Xin Ma, 27 Nov 2022
-
CC1: 'Comment on essd-2022-215', Mengqi Zhang, 09 Nov 2022
In this paper, a spatiotemporal interpolation method is developed, and a data product with full spatiotemporal coverage is generated by using the XCO2 data of GOSAT.I'm not particularly aware of the article category for ESSD, but compared to other research papers in ESSD, it's more suitable for technical description articles, at least at this stage. The article has some obvious scientific errors and inappropriate knowledge descriptions, the following points should be considered to improve the quality of the article, especially some major errors. As community comments, I believe these comments will increase the understanding of carbon monitoring satellite data assimilation and improve this research.
Although we are very concerned about the carbon cycle and the spatiotemporal distribution of CO2, for atmospheric inversion models, sparse data observations are sufficient to obtain carbon fluxes. NOTE I'm not denying that we don't need a spatially seamlessly CO2 distribution, but the introduction should explain why we need a spatiotemporally seamlessly CO2, such as calculating global averages, analyzing seasonal changes.
P6. Line 160. Equation 1 looks very strange and seems wrong. What is the difference between coefficients c and e? What is the difference between d and g? The result needs to be checked carefully and even recalculated. In addition, adopting this method to construct time series would lead to significant drawbacks. I believe the authors may not understand the CO2 growth rate, please see papers such as Buchwitz. et.al, 2018 and A. Chatterjee et.al, 2017 to understand the significance of CGR in reflecting vegetation, climate, etc. The method of presetting a function to fit will not be able to capture the real change in the CO2 growth rate, and the function will be directly known after derivation. One of the reasons we want seamless data is to better calculate the global average, and thus the growth rate, that is, the net flux. Fitting with a fixed function would loss this.
In the abstract and the introduction, the authors claim that the amount of XCO2 data is mainly affected by factors such as clouds and aerosols. This is wrong, it's actually a swath issue. The design of the width is related to the fluctuation amplitude of atmospheric CO2 and optical inversion factors. Authors are advised to read the relevant literature and correct the description in this section. While this is not that important for this article, the readers needs to understand the real background for this work.
P3, L65-83. This section overlaps with method descriptions in subsequent part, and it is not appropriate to introduce too much about the methods of this study in the introduction.
P4, L98 "...the accuracy of the comparison between the GOSAT data product and the TCCON site was 0.56 ppm", it is not appropriate to use the "accuracy" word, it should be stated, such as standard deviation, bias, etc.
P4, L100, "... three-day temporal resolution. The time resolution of GOSAT-2 satellite is 6 days..." is inappropriate. The correct description is the revisit cycle/repeat cycle. Please differentiate these concepts.(temporal resolution,time resolution, repeat cycle,)
P4, L105. However, the OCO-2_L2_Lite_FP9r provides data locations that are gradually shifted over time by satellite observations. This sentence is difficult to understand. I think you should express that the orbits of the sub-satellite points are evenly distributed? Illustration may be needed.
P4, L112 column-averaged XCO2. This is wrong. And the full name of XCO2 is wrong, including the title, abstract and etc. I think it shold be carefully checked the full text of the corresponding full scientific name.
P4. L107, "fixed location" should be more clear. L109. the six data channels are wrong. TANSO-FTS is a 4–band interferometer.
P4., L123. Please correct for column-averaged abundances of CO2 expression. And the results showed that R2 was 0.9686, and RMSE was 1.3811 . Please indicate the source.
P5. 146 EBK theory or EBK method ? Please express it in a unified way. such as L150.
In Section 2.4, more indicators for accuracy evaluation should be added. Bias and standard deviation are necessary in the verification of XCO2.
P7. 195. It is not necessary to show the results of each year, only a few examples, such as some months or a specific year, are sufficient. From the analysis of the data, such as some seasonal changes, changes in CO2 growth, and spatial differences may be more meaningful.
Figure 6, As said at the beginning, of course we know that CO2 is rising, but its growth rate is more meaningful, and it is recommended to draw a related graph of the growth rate. If it does not reflect reasonable fluctuations, but a fully sinusoidal pattern, the study would be significantly flawed.
From the research point of view, averaging kernel and the prior profile should be considered in comparison with OCO-2. Although they may be ignored in some cases and not important on monthly validation where accuracy is not required, the article should mention it.
Figure 3,4 and 5. PXCO2 TXCO2 P XCO2 should be described uniformly (note the space). Other than that, I would suggest that it would be better to do time series validation on a monthly basis. Specifically, the horizontal axis is time, and the vertical axis is parameters such as error, which can also be filled with error distribution, which is more intuitive.
Figure 7~17. It looks like this resolution may be trapped in a highly smooth phenomenon, which means it may not really be 0.25 degrees. It is recommended to draw a detailed map of some regions to show that the method does have this good resolution and can capture reasonable and sufficient spatial gradient changes. It is also recommended to compare the results of models, such as CarbonTracker or the L4B model products of GOSAT-NIES, to demonstrate the rationality of the results.
As mentioned above, in addition to the error evaluation in the time dimension, the error in the spatial dimension should also be evaluated to illustrate the reliability of the data.
I am not an Native English speaker. But I believe that the English of this article should be greatly improved, including scientific names of many nouns, and descriptions in scientific language.
Buchwitz, Michael, et al. "Computation and analysis of atmospheric carbon dioxide annual mean growth rates from satellite observations during 2003–2016." Atmospheric Chemistry and Physics 18.23 (2018): 17355-17370.
Chatterjee, A., et al. "Influence of El Niño on atmospheric CO2 over the tropical Pacific Ocean: Findings from NASA’s OCO-2 mission." Science 358.6360 (2017): eaam5776.
Citation: https://doi.org/10.5194/essd-2022-215-CC1 -
AC2: 'Reply on CC1', Xin Ma, 27 Nov 2022
Dear reviewer, thank you for your kindly suggestions. For specific responses, please refer to the attachment.
-
CC2: 'Reply on AC2', Mengqi Zhang, 27 Nov 2022
The authors put a lot of effort into addressing most of my concerns and suggestions. The authors address most of the errors and concerns, especially fixing Equation 1. The author added more validation and analysis content, which greatly improved the quality of the article.
As a community comment, overall, the quality of the article has been greatly improved. I think it meets the basic criteria for publishing on ESSD at least for now.
But some content still needs polishing. I have some suggestions as follows.
On the response of "P4, L105. However, the OCO-2_L2_Lite_FP9r provides data locations that are gradually shifted over
time by satellite observations", I already understand what the author wants to convey. But in fact, "the satellite orbit is gradually offset in order
to collect more data." is the Inherent Properties of Sun-Synchronous Orbits(SSO). Most polar-orbiting remote sensing satellites, such as Terra and Aqua, opt for SSO orbits, which allow them to image the globe in a single day. For OCO-2, it is also an SSO orbit, but its width is so narrow that it can only detect sub-satellite points (there are also SAM mode and glint mode). Here I just want to remind, not deny anything.If "Fig. 13-1 The distribution of seasonal mean XCO2 in 2010" is added to the revised manuscript, it is better to replace abcd with the name of the season, or the abbreviation of the month (eg JFM).
If "Figure 16-1 Monthly-averaged XCO2 validation results for TCCON data and CDC dataset at global mid- and low-latitude TCCON sites from 200906 to 202012. " is added to the revised manuscript, it is recommended to check the color fill of the standard deviation. Although readers can understand its meaning, 1x std and 2x std seem to be reversed, maybe it may be a color overlay problem, it is recommended to check, including others.
According to ESSD's publication policy (https://essd.copernicus.org/articles/10/2275/2018/, ), Acknowledgements should mention the OCO-2 and GOSAT data used in this article, as well as the TCCON data, and other publicly available data used. This is also a general requirement.
Overall, the author has solved most of the problems, and I believe the above comments can better improve the quality of the article. As a community comment, I have no further comments.
Citation: https://doi.org/10.5194/essd-2022-215-CC2
-
CC2: 'Reply on AC2', Mengqi Zhang, 27 Nov 2022
-
AC2: 'Reply on CC1', Xin Ma, 27 Nov 2022
- RC2: 'Comment on essd-2022-215', Anonymous Referee #2, 26 Nov 2022
Data sets
WHUXCO2-GLOBAL Haowei Zhang, Boming Liu, Xin Ma, Ge Han, Qinglin Yang, Yichi Zhang, Tianqi Shi, Jianye Yuan, Wanqi Zhong, Yanran Peng, Jingjing Xu, Wei Gong https://doi.org/10.6084/m9.figshare.17826404.v4
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,141 | 354 | 57 | 1,552 | 56 | 49 |
- HTML: 1,141
- PDF: 354
- XML: 57
- Total: 1,552
- BibTeX: 56
- EndNote: 49
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1