the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 24 Apr 2025
5 Years of GOSAT-2 Retrievals with RemoTeC: XCO2 and XCH4 Data Products with Quality Filtering by Machine Learning
Abstract. Accurately measuring greenhouse gas concentrations to identify regional sources and sinks is essential for effectively monitoring and mitigating their impact on the Earth’s changing climate. In this article we present the scientific data products of XCO2 and XCH4, retrieved with RemoTeC, from the Greenhouse Gases Observing Satellite-2 (GOSAT-2), which span a time range of five years. GOSAT-2 has the capability to measure total columns of CO2 and CH4 to the necessary requirements set by the Global Climate Observing System (GCOS), who define said requirements as accuracy < 10 ppb and < 0.5 ppm for XCH4 and XCO2 respectively, and stability of < 3 ppb yr−1 and < 0.5 ppm yr−1 for XCH4 and XCO2 respectively.
Central to the quality of the XCO2 and XCH4 datasets is the post-retrieval quality flagging step. Previous versions of RemoTeC products have relied on threshold filtering, flagging data using boundary conditions from a list of retrieval parameters. We present a novel quality filtering approach utilising a machine learning technique known as Random Forest Classifier (RFC) models. This method is developed under the European Space Agency’s (ESA) Climate Change Initiative+ (CCI+) program and applied to data from GOSAT-2. Data from the Total Carbon Column Observing Network (TCCON) are employed to train the RFC models, where retrievals are categorized as good or bad quality based on the bias between GOSAT-2 and TCCON measurements. TCCON is a global network of Fourier transform spectrometers that measure telluric absorption spectra at infrared wavelengths. It serves as the scientific community’s standard for validating satellite-derived XCO2 and XCH4 data. Our results demonstrate that the machine learning-based quality filtering achieves a significant improvement, with data yield increasing by up to 85 % and RMSE improving by up to 30 %, compared to traditional threshold-based filtering. Furthermore, inter-comparison with the TROPOspheric Monitoring Instrument (TROPOMI) indicates that the quality filtering RFC models generalise well to the full dataset, as the expected behaviour is reproduced on a global scale.
Low systematic biases are essential for extracting meaningful fluxes from satellite data products. Through TCCON validation we find that all data products are within the breakthrough bias requirements set, with RMSE for XCH4 <15 ppb and XCO2 <2 ppm. We derive station-to-station biases of 4.2 ppb and 0.5 ppm for XCH4 and XCO2 respectively, and linear drift of 0.6 ppb yr−1 and 0.2 ppm yr−1 for XCH4 and XCO2 respectively.
For XCH4, GOSAT-2 and TROPOMI are highly correlated with standard deviations less than 18 ppb and globally averaged biases close to 0 ppb. The inter-satellite bias between GOSAT and GOSAT-2 is significant, with an average global bias of -15 ppb. This is comparable to that seen between GOSAT and TROPOMI, consistent with our findings that GOSAT-2 and TROPOMI are in close agreement.
Competing interests: Three of the co-authors are members of the editorial board for Atmospheric Measurement Techniques in the subject area of Gases
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(8991 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-3990', Robert Parker, 02 Jul 2025
Reivew of Barr et al., 5 Years of GOSAT-2 Retrievals with RemoTeC: XCO2 and XCH4 Data Products with Quality Filtering by Machine Learning
Here the authors present new GOSAT-2 datasets based on the RemoTeC algorithm for XCO2 and XCH4. They also deploy a new method for post-retrieval quality filtering that substantially improves the data yield.
This is a very nice study that should help advance how we handle the data quality of these sorts of retrievals in the future, with the method generally applicable to future missions . Overall, I recommend this manuscript for publication after addressing the following comments. These are ordered by line-number and the majority just seek additional clarification, justification or discussion to be added to the text.
L12 - Do TCCON/GOSAT-2 co-locations sample an adequate range of geophysical parameters (aerosols, albedos, etc) to produce a robust post-filter? Many approaches use multiple “truth metrics” (e.g. TCCON, models, small area approximation, etc). You deliberately include some high-albedo cases into your training data to compensate for this but do I understand correctly that you do not have TCCON co-locations for these? How confident are you that this data is not biased differently?
L24 - Why do GOSAT and GOSAT-2 disagree (with the same algorithm and priors?) compared to TROPOMI?
L31 - Some minor typos should be corrected, e.g. “anomoly”.
L72 - Is the stated TCCON performance (L72) relevant for GGG2020 or does it relate to older GGG2024 data?
L83 - Rather than footnotes for the ESA CCI documents (L83/84), can they please be cited fully in the bibliography?
L87 - The instrument line shape for GOSAT-2 potentially has caused some issues with other retrieval algorithms. Could you please elaborate on its usage here (L87) and whether, if at all, you do anything to compensate for these (e.g. fitting a shift/stretch). I would also strongly recommend including a figure of a typical spectral residual for each species so the quality of the fit can be shown.
L147 – Would it be possible to outline your full state vector (or explicitly link back to section/table in previous work where this is fully described)?
Table 1 – Can it please be made explicit in the table which of these do not apply to the Proxy (i.e. 6-8?).
L157 - Can you specify which TCCON version is used for this study (and ideally the temporal extent of the different TCCON datasets)?
L158 – Given that the XCH4 FP and Proxy (post-bias correction) have quite different biases, can you also compare the non-bias corrected data? Is this difference in bias coming from the original data or from the correction?
L201 - Can you elaborate further how the value of XT is decided upon?
L211 - Does taking the different filtering approaches for land vs glint (L211) lead to significant differences in the sampling statistics? Could this lead to ocean/land biases in the final data? It would also be good to see how the data looks for a few single orbits that pass over both ocean/land.
L217 – How true is that assumption and is it robust to issues such as sensor degradation (that we know GOSAT/GOSAT-2 can suffer from) which have a strong temporal component.
L237 – This may not be easily possible but it would be very interesting if the models for the different years themselves were all similar. They all clearly give consistent results but are the models compensating for different things in different years, to varying degrees. Some examination of the potential explainability of these models would be great but may not be possible.
L244 – This seems to be saying that as there’s typically much less “bad” Proxy data, it is harder to identify the bad cases as they stand out less. Could you evaluate the model CO2 used in the proxy in a similar way here to separate the components?
L259 – Can you elaborate on why XCH4 and XCO2 FP (from the same retrieval?) have different yields?
L269 – Lots of mentions of TCCON prior to defining that it’s GGG2020 that is used. I’d mention this sooner.
Figure 3 – Am I correct that this mixes together land and glint data? Can you separate the two out for some sites to better understand any ocean/land bias in the data?
L304 – Is this a fair comparison between GOSAT and GOSAT-2 when you have applied this new post-filtering and hence can define/tune the RMSE for GOSAT-2?
L316 – Minor grammar issue “by in the ratio”
L337 – How are you matching GOSAT to GOSAT-2 data? i.e. What criteria do you use to find co-located soundings?
Figure 7 – Can you show similar maps for the other 2 products? Maybe in an appendix?
L381 – Can you outline how you match GOSAT-2 to TROPOMI?
L387 – I don’t quite understand this point about the bias remaining “effectively constant”. The bias increases with QA value doesn’t it? (i.e. from -4.6 to -6.3 ppb). Can you also comment on why the proxy bias seems to systematically decrease with increasing QA?
L390 – typo (missing word after Northern?)
Citation: https://doi.org/10.5194/egusphere-2024-3990-RC1 -
AC2: 'Reply on RC1', Andrew Barr, 03 Sep 2025
To the editor:
We thank the referee for their kind words and constructive feedback. To improve the manuscript accordingly we will incorporate each of the points raised in the revised paper. We give responses to each comment below. Comments from the referee are given followed by our response. We give line numbers and Table/Figure numbers according to the revised manuscript.
L12 - Do TCCON/GOSAT-2 co-locations sample an adequate range of geophysical parameters (aerosols, albedos, etc) to produce a robust post-filter? Many approaches use multiple “truth metrics” (e.g. TCCON, models, small area approximation, etc). You deliberately include some high-albedo cases into your training data to compensate for this but do I understand correctly that you do not have TCCON co-locations for these? How confident are you that this data is not biased differently?
The referee raises here an important point about the data used for training the machine learning filtering models. There are two questions asked here: 1) Is an adequate parameter space covered during training of the models? and 2) Does the different labelling of training data, depending on the albedo, lead to different results in the final product? We answer these both one by one.
i) Figure 2 has been expanded to present the ranges of a number of variables - including albedo, solar zenith angle, surface elevation variance among others - for high quality examples only. Here we choose the 2019 filtering model for the XCH4 Full Physics product. The combined training dataset sufficiently covers the expected ranges of these geophysical parameters. We discuss these additional panels in the text at Lines 226 – 230.
ii) Due to the coverage of stations, training data from TCCON alone do not cover the full albedo range in the GOSAT-2 data. To avoid potentially biasing the filtering models to lower albedo, training data were supplemented with a differently defined sample of data that includes soundings where the surface albedo is higher than 0.4. The additional training subset cannot be labelled in the same way as the TCCON subset, therefore we label it according to Table 1, and also cannot be validated in the same way either. To make sure there is not a significant different between the filtered data products based on the albedo, we have included an additional analysis between GOSAT-2 and TROPOMI, for albedo >0.4 only, and discuss this at Line 431. For estimating the bias, we exclude North Africa from the comparison due to the large bias in this region apparent in Figures 11 and 12. We find that the average bias of this sub-population is 2.1 ppb for the Proxy product and -2.6 ppb for the Full Physics product, which are in close agreement with that from Table 4. The manuscript has been updated accordingly.
L24 - Why do GOSAT and GOSAT-2 disagree (with the same algorithm and priors?) compared to TROPOMI?
The literature does not provide a clear conclusion on this point. In this paper we consider the longest timeseries in the literature, however the magnitude of the difference between GOSAT and GOSAT-2 level 2 products is partially shrouded by the different quality filtering methods applied (threshold filtering for GOSAT and RFC filtering for GOSAT-2). It could be fortuitous that GOSAT-2 and TROPOMI agree equally bad compared to GOSAT, however given the improved quality filtering of GOSAT-2 we state in the text (L439) that the disagreement between GOSAT and GOSAT-2 is due to biases in the GOSAT products.
L31 - Some minor typos should be corrected, e.g. “anomoly”.
Typo has been corrected.
L72 - Is the stated TCCON performance (L72) relevant for GGG2020 or does it relate to older GGG2024 data?
The performance stated is for the GGG2014 release. This has been clarified in the text (L80).
L83 - Rather than footnotes for the ESA CCI documents (L83/84), can they please be cited fully in the bibliography?
Full citations added for both PUGs and ATBDs for the Proxy and Full Physics products.
L87 - The instrument line shape for GOSAT-2 potentially has caused some issues with other retrieval algorithms. Could you please elaborate on its usage here (L87) and whether, if at all, you do anything to compensate for these (e.g. fitting a shift/stretch). I would also strongly recommend including a figure of a typical spectral residual for each species so the quality of the fit can be shown.
No stretch/squeeze or shift quantities are retrieved or fit to the ILS. We simply use it as is. A statement on this has been added to the text (L154). Following the referee’s recommendation, we have included an example of spectral fits per band for the Physics retrieval in Figure 1, which shows the measurement, model and residuals for a high quality retrieval. L119 cites this figure.
L147 – Would it be possible to outline your full state vector (or explicitly link back to section/table in previous work where this is fully described)?
The state vectors are fully described in the CCI+ ATBDs. We have given the citation and exact location of where it can be found at L154 and L159.
Table 1 – Can it please be made explicit in the table which of these do not apply to the Proxy (i.e. 6-8?).
We have now marked filters that do not apply to the Proxy product with an asterisk and also clarify this in the caption of Table 1.
L157 - Can you specify which TCCON version is used for this study (and ideally the temporal extent of the different TCCON datasets)?
Clarification that we use the GGG2020 release added at L170. The temporal extent at each site has been added to Table B1.
L158 – Given that the XCH4 FP and Proxy (post-bias correction) have quite different biases, can you also compare the non-bias corrected data? Is this difference in bias coming from the original data or from the correction?
I believe the question that is being asked here is how the Proxy and Full Physics XCH4 products compare before and after bias correction. To compare this we can look at the TCCON validation for the data before bias correction. We believe that a full TCCON validation of non-bias-corrected data would be confusing for the reader, and unconstructive for the paper given the already lengthy TCCON validation. We therefore limit the discussion here to mean bias over all TCCON colocations.
Before bias correction the mean bias is 7.2 ppb and 13.3 ppb for the Full Physics and Proxy XCH4 products, respectively. These change to -0.1 ppb and -0.1 ppb after bias correction. So the average bias is fit to approximately zero. The RMSE is also slightly improved from bias correction, but the change is minor, about 0.5 ppb. Thus before bias correction, the Full Physics approach is closer to TCCON than the Proxy, as expected. We have added a discussion on this in the manuscript (L326). We have also included some additional information on how the bias correction performs on the Proxy and Full Physics products in section 3.4 (L176).
L201 - Can you elaborate further how the value of XT is decided upon?
XT is chosen to probe the steepest part of the curves in Figure 3 thus maximising the improvement that can be extracted from the machine learning filtering approach. We have added a statement on this to the
L211 - Does taking the different filtering approaches for land vs glint (L211) lead to significant differences in the sampling statistics? Could this lead to ocean/land biases in the final data? It would also be good to see how the data looks for a few single orbits that pass over both ocean/land.
In order to address the referee’s points here, we have expanded section Appendix A in which ocean measurements are discussed, to include a description of the bias correction for glint mode, and added Figure A1 which shows 6 consecutive days of GOSAT-2 data - which is the revisit time of GOSAT-2. We also include a discussion on land/ocean biases (L486) revolving around the newly added Table A1, in which we quote the bias, per product, for TCCON stations that have colocated measurements over both land and ocean. Some stations used for validation of glint mode have very few colocations (1 or 2), so the significance of the statistics here should be taken into consideration when interpreting these conclusions. These stations are marked with an asterisk in Table A1. We find that the agreement between land/ocean is better for the Full Physics products, however the Proxy product shows some significant differences. This is explicitly stated in the manuscript (L491).
L217 – How true is that assumption and is it robust to issues such as sensor degradation (that we know GOSAT/GOSAT-2 can suffer from) which have a strong temporal component.
The machine learning models learn the relationship between a set of retrieval features and the quality of a measurement, which is defined as the difference with TCCON. Assuming that detector degradation leads to an increase in noise, and such an increase leads to a poorer quality retrieval, measurements suffering from such degradation should have a larger difference with TCCON and therefore be more likely to be flagged as bad quality. Looking at the feature importance analysis in Figure 2, the models across the different years are very similar, implying that this assumption is robust. We mention this briefly in the text (L242).
L237 – This may not be easily possible but it would be very interesting if the models for the different years themselves were all similar. They all clearly give consistent results but are the models compensating for different things in different years, to varying degrees. Some examination of the potential explainability of these models would be great but may not be possible.
L244 – This seems to be saying that as there’s typically much less “bad” Proxy data, it is harder to identify the bad cases as they stand out less. Could you evaluate the model CO2 used in the proxy in a similar way here to separate the components?
.
L259 – Can you elaborate on why XCH4 and XCO2 FP (from the same retrieval?) have different yields?
We clarify this point in L289.
L269 – Lots of mentions of TCCON prior to defining that it’s GGG2020 that is used. I’d mention this sooner.
An extra reference to that fact that TCCON GGG2020 is used throughout has been added at L221.
Figure 3 – Am I correct that this mixes together land and glint data? Can you separate the two out for some sites to better understand any ocean/land bias in the data?
Figure 4 shows the TCCON validation for land retrievals only. We have added this detail to the caption and L305 states this as well. The results for TCCON validation of ocean retrievals are presented seperately in Figure A2.
L304 – Is this a fair comparison between GOSAT and GOSAT-2 when you have applied this new post-filtering and hence can define/tune the RMSE for GOSAT-2?
The tuneability of the RMSE is one of the strengths of the new QA value presented for GOSAT-2. It should be noted that the comparison here is between data products and we refrain from commenting on whether GOSAT or GOSAT-2 performs better as a mission. In that sense we believe the comparison is fair. To make this more clear, we have removed this paragraph from section 5.1 so that the TCCON validation section deals only with GOSAT-2, and placed it at the end of section 5.2, where we compare the data products with their GOSAT counterparts. Here we also state that we are comparing data products of GOSAT and GOSAT-2 (L393).
L316 – Minor grammar issue “by in the ratio”
typo corrected
L337 – How are you matching GOSAT to GOSAT-2 data? i.e. What criteria do you use to find co-located soundings?
This has been added to the text (L363).
Figure 7 – Can you show similar maps for the other 2 products? Maybe in an appendix?
Similar maps of the XCO2 Full Physics and XCH4 Proxy products have been added in Figure C1 in a section in the Appendix. Reference is made to these figures at L369.
L381 – Can you outline how you match GOSAT-2 to TROPOMI?
The data from the two satellites are matched by re-gridding XCH4 to 2x2 degree lat/lon boxes, per day. A colocation is considered when there are data from each satellite in the same grid cell for a given day. We have added this to the text (L406). The colocation method is the same as for the GOSAT intercomparison, therefore to avoid repetition we refer to that part of the paper.
L387 – I don’t quite understand this point about the bias remaining “effectively constant”. The bias increases with QA value doesn’t it? (i.e. from -4.6 to -6.3 ppb). Can you also comment on why the proxy bias seems to systematically decrease with increasing QA?
We thank the referee for this point, it has helped to clarify an important conclusion of the paper. We have decoupled the statement about the bias from the start of this paragraph as the key result is more that sigma_TROPOMI follows sigma_TCCON with QA value.
We have noted the systematic change in bias for the Full Physics product, however the change for the Proxy is 0.7 ppb. If 18 ppb is taken as 1 % of XCH4, this would correspond to 0.04 %, a change of which we are not sensitive to, so interpretation of this should be taken with caution. Both of these points have been added to the manuscript (L424).
L390 – typo (missing word after Northern?)
This indeed was a mistake in the wording and has now been fixed.
Citation: https://doi.org/10.5194/egusphere-2024-3990-AC2
-
AC2: 'Reply on RC1', Andrew Barr, 03 Sep 2025
-
RC2: 'Comment on egusphere-2024-3990', Gregory Osterman, 29 Jul 2025
I apologize for being late with this feedback.
This paper uses a machine learning to develop a new way of quality filtering retrievals of column mixing ratios of carbon dioxide (XCO2) and methane (XCH4) from the GOSAT-2 satellite. The authors show that an improved data product can be developed from the GOSAT-2 data retrievals using the RemoTeC processing and a Random Forest Classifier model. The authors present the details of the GOSAT-2 retrievals, the model used in filtering the data and the data product results. They then use data from TCCON to validate their results and show comparisons between TCCON and their data set for XCO2 and XCH4. They also show data intercomparisons between GOSAT-2 and GOSAT retrievals using RemoTeC. Finally for XCH4 they show comparisons from their GOSAT-2 data set and TROPOMI. The work is done on a five year data record.
The results they obtain show they have developed a QA process that allows users to adjust parameters to provide more “good” quality data, though this will come at the expense of the errors in comparison to TCCON. The authors also explain well how TCCON data can be used in both the quality control process and also for validation.
This is a well written manuscript, with a well thought out approach and presentation. I would recommend it for publication. The results are compelling for their selected subset of the GOSAT-2 data and improving the satellite data record. The techniques utilized in this analysis could potentially be used with other satellite data records to improve the quality and flexibility of use of the products.
I do have a few minor comments, listed below.
Comments:
Line 31: anomaly is misspelled
Line 31: maybe “increase” instead of “excess”
Line 32: preindustrial levels
Line 43: classified instead of “classed”
Lines 53 and 60: Authors mention level-1B and level 2 without really saying what that means, though they do mention radiance spectra. Maybe that could be clarified?
Line 65: Could you provide a little bit more information it the text as to why “Science with TANSO-FTS-2 has been limited, and more restricted to total column products”. You have the references, but it could be helpful to provide a brief mention of why that is the case so readers do not need to go the references for context.
Line 70: “has for long been widespread used”
Line 73: TCCON uses different methods of obtaining vertical profiles at their sites, maybe mention balloon-borne observations (AirCore) in addition to aircraft. Also maybe mention that these are site specific and can vary quite a bit.
Lines 258-265: The authors highlight the results from the RFC model of quality controlling the GOSAT-2 data. Figure 2 includes single data points showing what data filtering using table 1 would look like. But that is not mentioned in the text. What would the results for ocean data look like from table 1?
Figure 3: The authors mention on line 276 that “These are single soundings of GOSAT-2 over land compared to an average of the TCCON measurements that coincide spatially and
temporally”, is that the reason for the lines of (along the x-axis) of data that appear in the scatter plot. What would a daily average of both satellite and TCCON look like? Would there be much difference?
Line 309: Why are there a disproportionately higher number of collocations at Caltech, Edwards and Xianghe? Maybe a little more explanation on that?
Section 5.2: I was not clear on this, maybe I missed it, but did you use the RCF on the GOSAT data (I think you did not). How is the GOSAT data quality filtered? Is it similar to the Table 1 approach? What changes with the comparison if you use Table 1 and compare to GOSAT?
Line 375/Table 4: Again, I apologize if I missed this, but do you recalculate the TCCON comparisons to include data that has GOSAT-2/TROPOMI comparisons? Table 4 are the global numbers for different QA values. Column 6 is the scatter for a TCCON comparison on just the GOSAT-2/TROPOMI subset? That number is higher than Table 3?
Line 390-392: The wording here is confusing, maybe rephrase this to make clear that there is a large difference over North Africa and that they could be attributed to high aerosol levels from dust and burning events causing issues for the retrievals.
Line 396: “The global average is close to zero”, I think you mean that the aggregate global difference between TROPOMI and GOSAT-2 is close to zero?
Could this data filtering model be used on GOSAT? Would that make the data sets from the two satellites more compatible?
Citation: https://doi.org/10.5194/egusphere-2024-3990-RC2 -
AC1: 'Reply on RC2', Andrew Barr, 03 Sep 2025
To the editor:
We thank the referee for their kind words and constructive feedback. To improve the manuscript accordingly we will incorporate each of the points raised in the revised paper. We give responses to each comment below. Comments from the referee are given followed by our response. We give line numbers according to the revised manuscript.
Line 31: anomaly is misspelled. maybe “increase” instead of “excess”
Line 31: typo corrected and “excess” changed to “increase”
Line 43: classified instead of “classed”
Line 43: “classed” changed to “classified”
Lines 53 and 60: Authors mention level-1B and level 2 without really saying what that means, though they do mention radiance spectra. Maybe that could be clarified?
We have slightly reordered the paragraph that introduces GOSAT to expand the definition of L1B data. Also the referee is correct, since reference is repeatedly made to level 2 data throughout the paper, this deserves its own definition. We give this at Line 62.
Line 65: Could you provide a little bit more information it the text as to why “Science with TANSO-FTS-2 has been limited, and more restricted to total column products”. You have the references, but it could be helpful to provide a brief mention of why that is the case so readers do not need to go the references for context.
Line 68: Several interesting science studies are worth noting here, therefore we have decided to rework this paragraph. Particularly the very recently submitted work of Janardanan et al. comparing flux inversions of CH4, when assimilating GOSAT or GOSAT-2 XCH4 data, and the retrieval of HDO/H2O ration from combined NIR and TIR bands. We also introduce here the level 2 products of NIES and IUP-Bremen, rather than do this in section 5.1, as it makes the flow of the discussion better.
Line 70: “has for long been widespread used”
Line 77: phrase changed
Line 73: TCCON uses different methods of obtaining vertical profiles at their sites, maybe mention balloon-borne observations (AirCore) in addition to aircraft. Also maybe mention that these are site specific and can vary quite a bit.
Line 80: reference to AirCore added along with Karion et al. 2010.
Lines 258-265: The authors highlight the results from the RFC model of quality controlling the GOSAT-2 data. Figure 2 includes single data points showing what data filtering using table 1 would look like. But that is not mentioned in the text. What would the results for ocean data look like from table 1?
Reference to data filtering using Table 1 in comparison to the RFC filtering was made already (see Line 286). For Figure 3, the number of collocated measurements over ocean is so small compared to land (factor of 100), that a data point for ocean would look like it lies along the x-axis. We therefore made no update to the manuscript.
Figure 3: The authors mention on line 276 that “These are single soundings of GOSAT-2 over land compared to an average of the TCCON measurements that coincide spatially and temporally”, is that the reason for the lines of (along the x-axis) of data that appear in the scatter plot. What would a daily average of both satellite and TCCON look like? Would there be much difference?
Figure 4: the referee is correct, lines of data in the x-axis direction result from taking the daily averaged TCCON value plotted against single soundings of GOSAT-2, and are indicative of a bias over geolocation. A daily average of GOSAT-2 data would reduce the overall scatter considerably. GCOS however define the precision in terms of single soundings. We address this point in Line 308.
Line 309: Why are there a disproportionately higher number of collocations at Caltech, Edwards and Xianghe? Maybe a little more explanation on that?
At Line 340, we have expanded a bit more this discussion to explain why there are more colocations at Edwards and Caltech, and the impact of this on the results.
Section 5.2: I was not clear on this, maybe I missed it, but did you use the RCF on the GOSAT data (I think you did not). How is the GOSAT data quality filtered? Is it similar to the Table 1 approach? What changes with the comparison if you use Table 1 and compare to GOSAT?
There are two points made by the reviewer here 1) How are the GOSAT data quality filtered and 2) how do the results change when comparing GOSAT and GOSAT-2 both filtered with Table 1. We shall answer these both respectively. i) Indeed the RFC filtering has not been applied on GOSAT data. The GOSAT data are filtered according to Table 1. The referee here raises an important point that has been missed. We have therefore included a sentence at the start of section 5.2 (Line 364) to highlight how the filtering of GOSAT is done. ii) If GOSAT-2 is filtered according to Table 1 (therefore the same as GOSAT) the bias is reduced for the Full Physics XCH4 product, which goes from -15 ppb to -9.3 ppb. For the Proxy XCH4 product the effect is less where the bias changes from -5.3 ppb to -6.5 ppb. To avoid confusion for the reader as to which results are presented for which quality filtering, we do not include these values in the manuscript.
Line 375/Table 4: Again, I apologize if I missed this, but do you recalculate the TCCON comparisons to include data that has GOSAT-2/TROPOMI comparisons? Table 4 are the global numbers for different QA values. Column 6 is the scatter for a TCCON comparison on just the GOSAT-2/TROPOMI subset? That number is higher than Table 3?
The results presented in Table 4 are for the TCCON validation for all GOSAT-2/TCCON colocations. This is now explicitly stated in Line 423. The referee makes a good point here that Table 4 should be consistent with Table 3. This was not the case due to a different way of defining sigma_TCCON. To avoid confusion we have recalculated the values for sigma_TCCON in Table 4 as the mean RMSE over all station, as done in Table 3. We also state this in Line 424.
Line 390-392: The wording here is confusing, maybe rephrase this to make clear that there is a large difference over North Africa and that they could be attributed to high aerosol levels from dust and burning events causing issues for the retrievals.
Line 427: There was a mistake made with this sentence. It has been reworded
Line 396: “The global average is close to zero”, I think you mean that the aggregate global difference between TROPOMI and GOSAT-2 is close to zero?
Line 437: Wording has been clarified
Could this data filtering model be used on GOSAT? Would that make the data sets from the two satellites more compatible?
The RFC filtering could indeed be applied to GOSAT, and this would make for a better comparison allowing one to more concretely comment on the performance of GOSAT vs GOSAT-2. We consider this out of the scope of the paper, which presents the GOSAT-2 data products, and would be more suitable for a follow-up paper. We address this point in Line 395 of the manuscript.
Citation: https://doi.org/10.5194/egusphere-2024-3990-AC1
-
AC1: 'Reply on RC2', Andrew Barr, 03 Sep 2025
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 551 | 83 | 21 | 655 | 24 | 52 |
- HTML: 551
- PDF: 83
- XML: 21
- Total: 655
- BibTeX: 24
- EndNote: 52
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1