the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 10 Apr 2026
Vicarious calibration of the JAXA Earth Clouds, Aerosols and Radiation Explorer (EarthCARE)/Multi-Spectral Imager (MSI) level 2 cloud product (MSI_CLP) and aerosol product (MSI_ARL)
Abstract. After launch in May 2024, the Earth Clouds, Aerosols and Radiation Explorer (EarthCARE) satellite entered its orbit and commenced operations. In the early studies, it was reported that the initial version of Multi-Spectral Imager (MSI) level 1 data (MSI_RGR) had suffered some issues that compromised the data quality. To solve the issues of the MSI level 1 data, vicarious calibration is necessary particularly for the visible, near infrared, and shortwave infrared (VNS) bands. This paper described the vicarious calibration on MSI_RGR using the geostationary satellite Himawari-9 cloud retrieval as an intermediator. Our results based on the late revised version (vBa) of MSI_RGR verified the data quality improvement work by the ESA MSI level 1 team, which from slight overestimation (calibration coefficient of about 0.93) for the visible band and almost no overestimation or underestimation for near-infrared band in the case of cloud product. On the other side, for aerosol product, slight overestimation (calibration coefficient of about 0.92–0.93) for the visible band and slight underestimation (calibration coefficient of about 1.04–1.07) for near-infrared band. The calibration coefficients derived here will be used for further development of EarthCARE MSI products.
- Preprint
(2457 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-1679', Anonymous Referee #1, 04 May 2026
-
RC2: 'Comment on egusphere-2026-1679', Anonymous Referee #2, 13 May 2026
General comment:
The manuscript by Wang et al. introduces a vicarious calibration concept for MSI Level-1 radiances using the Himawari-9 cloud and aerosol retrieval as intermediator to derive Level-1 calibration coefficients through radiative transfer calculations using the MSI LUTs. In contrast to vicarious calibration using pseudo-invariant calibration sites, this method presents an interesting alternative which allows fast accumulation of big data samples, because inter-satellite viewing geometry differences are relatively negligible.
The manuscript shows the improvements using a few case studies and corresponding improvements, but it lacks additional discussion, conclusions and statistics demonstrating the impact of the calibration coefficients on the agreement of cloud and aerosol products between MSI and Himawari. The paper will provide an important reference for future MSI updates and for the calibration of other satellite imagers, but needs major updates to also demonstrate this to the reader.
Detailed comments:
Your method targets liquid water clouds and you apply the calibration factor not to improve the MSI L1 M-RGR product, but “on-the-fly” when the MSI_CLP retrieval reads the M_RGR data. But how about ice clouds? I think that this might be wrong for the SWIR-2 band due to strong absorption of ice particles. I mean, your findings suggest that the SWIR-2 calibration update is very small anyway, but maybe not negligible. Also, the MSI_ARL vicarious calibration method suggests that SWIR-2 and SWIR-1 bands are underestimated. So, I guess, it will have an effect on the ice cloud retrievals too.
You are assuming the Himawari-9 cloud product as the truth, but you are not citing an article showing the Himawari-9 radiometric performance. Can you add in the introduction some literature showing “how good” Himawari is? Also, are there studies validating the Himawari cloud products? Since your methods relies on Himawari, it is required to provide some information about the associated uncertainties.
It would be interesting to see the agreement between Himawari and MSI COT and CRE before and after the vicarious calibration. Same also for AOT. You provide the scatter plots Fig 9-11 for the BA versus simulated to show the fit as correction coefficient and visually in Fig. 14 the observed and simulated AOT. But I miss some kind of quantification of the impact of the vicarious calibration on the aerosol and cloud product. Since you are using the same algorithms as Himawari-9 for both, it would be logical to demonstrate the improve fit between COT and CRE and AOT between MSI and Himawari before and after the vicarious calibration.
I would highly recommend to combine Fig. 4 with Fig. 6 and Fig.5 with Fig. 7. Right now, it makes it difficult for the reader to visually compare both products. I would suggest to have for both frames a plot consisting of three. The MSI product on the left, Himawari-9 in the middle, and on the right the Himawari-9 with the MSI frame included/overlayed. Or alternatively, plot a rectangle in the Himawari-9 scene showing the MSI swath. Anyway, having them directly next to each other improves the reading.
Fig.9 - Fig.11: Would it make sense to include all pixel in one plot and just say the random error as measure of the spread? I think that there is only a small added value of showing all three scatter plots for all three cases, while the case study overview above is very interesting.
Many figures are not well described in the manuscript, and the details are also not provided in the figure caption. Please add descriptions and scientific meanings for all figures so that the reader is able to carry a key message from each of the figures.
Selection of minor technical comments (all of them and additional spelling errors or consistency suggestions can be found in the annotated PDF - see the supplement):
- 1: I would suggest to shorten the title and delete the shortcuts, which you introduce in the text. Just something like: "Vicarious Calibration of the EarthCARE MSI JAXA Level-2 cloud and aerosol products."
- L13-14: “In the early studies, it was reported that the initial version of Multi-Spectral Imager (MSI) level 1 data (MSI_RGR) had suffered some issues that compromised the data quality”: please specify what issues the data suffered of. And specify that you are addressing the radiometric issues not the geolocation and coregistration. Maybe better say that early results showed offsets of the MSI radiances relative to Himawari radiances. This is more concrete than "suffered issues".
- 18: Unclear what is meant here. this overestimation is seen in your results before or after the corrections by the L1 team?
- 27: cloud-aerosol-radiation interactions.
- 32: The JAXA EarthCARE MSI Level-2 cloud product (MSI_CLP) has been ...
- 33: In addition, the JAXA EarthCARE MSI Level-2 aerosol- ...
- 34: For completeness, please also mention the European Level-2 products here (Hünerbein et al. and Docter et al.)
- 44: Muto et al. 2025: I cannot find the reference under references.
- 46: Applying vicarious calibration can help mitigating the issues observed in MSI Level-1 data.
- 50: “for late revised version BA”: for later application in the version vBA Level-2 processors MSI_CLP and MSI_ARL. (the reader should not get confused here whether the coefficients are applied to Level-1 MSI_RGR data, which is not the case.)
- 57: “problems in the initial…”: Please also specify here better, which problems. In absolute differences? Or in solar irradiance?
- 58: It would be good to make a statement here that radiometric calibration is usually done on board, but because there were issues observed with the measured solar irradiance and offsets observed in validation of the Level-1 data compared to other satellite sensors (Himawari, SEVIRI...), it was decided that ad-hoc vicarious calibration is required.
- 69: “clear confidence level”: Do you mean "cloud mask" with clear confidence level? I think that this is more common to say, or "cloud_flag".
- 76: “expected to be overestimated”: But CER is anticorrelated to the SWIR band. If SWIR is overestimated, CER will be underestimated (less absorption leads to smaller ice crystals or droplets).
- 77: “released on 27 May…”: You could cite here the Level-1 data disclaimer: https://earthcarehandbook.earth.esa.int/documents/d/earthcare-data-handbook/earthcare-msi-level-1-product-disclaimer
- 79: “Frame E data”: Could you give a brief overview which frames exist and how they are defined? For example, here: https://earthcarehandbook.earth.esa.int/article/product
- 81: “from 22 September 2024”: Does this mean that you use data from before baseline BA? Or the reprocessed BA, which includes September 2022 already? It would be more consistent, if your analysis already builds on the dataset that has been improved by ESA through BA. I think that you use BA, but from reading, it looks like you are using data from before. You can resolve this by just saying that BA was reprocessed covering the entire mission period dating back to 16 July 2024.
- 81: “high solar elevation results…”: Is this a criterion? Do you want high solar elevation? If yes, please also explain the choice of your QC.
- 93: For MSI_CLP, you describe how the Level-1 data issues influence the L2 cloud products. Could you also mention how Level-1 is expected to influence MSI_ARL?
- 95: “smile effect”: I would suggest to cite here the two SMILE papers: https://amt.copernicus.org/articles/16/603/2023/ (your own) and https://doi.org/10.5194/amt-17-2507-2024 (Docter et al. )
- 112: Is this a criterion for the vicarious calibration? If yes, I would suggest to include this in the method section (2.5)
- 112-114: The whole paragraph matches better in the concept of the vicarius calibration.
- 114-116: I would suggest to move this part in the conclusions/outlook.
- 146: “…were normalized…”: Please specify what normalization method was used.
- 151-160: This would better fit to the introduction, motivating why your method is somhow more robust for water clouds.
- 176: Is a similar QC also used for the water cloud vicarious calibration?
- 184-190: Part of this describe the method, QC and case selection. You could add another subsection to section 2, saying something like "Data sample selection" or "Data filtering".
- 9-Fig.11: My impression is that you remove outliers to strongly. Yes, for a stable fit, it is important to exclude extreme outliers, but you are cutting all variability too. Also, at the top of the cone, there is a bright line, which indicates that there is a high density of data "hidden" behind. I would relax the outlier screening a bit.
- 247-250: This doubles with the concept description.
- 255-256: This is an important result. I would suggest adding a table that shows the bands, and associated calibration coefficients. And I would suggest to add the coefficients from both methods in the table, also indicating which method is used.
- 12: You mention the red square in Fig 14 but not here. As the reading, I also want to know what the square means here in Fig. 12. Any chance of improving the image? It is hard to see anything. Maybe show only a subset or remove the titles above the figures, to reduce the white space between the figures. And you could also limit the max values in the colorbar to show more contrast.
- 14: Here, I read from the figure, that the AOT has increased for the simulated ref. Does this come from the underestimation of the SWIR bands? Adding scatter plots showing the AOT before and after the update would help demonstration of the improvements of the results, also in comparison to Himawari.
- Conclusions: Need to work on the conclusions. It is too short and too little discussion points. What are the limitations? What would you advise to do next? Does this kind of methods need to continue throughout the mission or do you think that this one time calibration correction solves it for the Level-2 products now for the entire time?
- 284: “was confirmed for the visible band”: which is in good agreement with the other method using water clouds.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 241 | 50 | 16 | 307 | 27 | 30 |
- HTML: 241
- PDF: 50
- XML: 16
- Total: 307
- BibTeX: 27
- EndNote: 30
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript aims to derive vicarious calibration coefficients for EarthCARE/MSI radiance with the purpose of improving the JAXA MSI level-2 cloud product and aerosol product. For the cloud product, the authors use Himawari-9 retrieved cloud optical thickness and cloud effective radius as intermediate cloud properties, then use the MSI cloud LUT to simulate the corresponding MSI radiance and compare it with the observed MSI radiance. Similarly to aerosol products.
The general concept of the method is understandable, and I do not see an obvious fundamental logical problem in using a well-characterized external satellite product as an intermediary for vicarious calibration. However, the current manuscript is not yet in good shape. The main problems are not only language and figure quality, but also the lack of deeper analysis and weak presentation of the results. In many places, the manuscript simply shows a figure and states that the result looks reasonable, without providing enough quantitative analysis, uncertainty discussion, or clear logic connecting the figure to the conclusion.
Specific comments
1. The manuscript lacks in-depth analysis. The current Results and Discussion section is mostly descriptive. The authors show figures and state that the simulated and observed radiances or reflectances agree reasonably well, but there is little detailed analysis or quantitative discussion.
2. Almost all figures need to be redesigned. Many figures are too small, lack readable labels, and do not include subpanel indices. Some figures appear to be included mainly to show that a product was generated, rather than to support a scientific point. Figure captions often lack key information needed to understand how the figure was produced. The authors should carefully decide which figures are necessary and redesign the remaining figures to directly support the manuscript’s arguments.
3. The schematic in Figure 1 does not clearly show the calibration procedure. I understand that the authors may want to illustrate the relationship between Himawari and MSI observations, but the current design is not scientifically informative. The decorative background image makes the figure visually busy without clarifying the method. Since Figure 2 already covers much of the same conceptual content, Figure 1 could be removed, or it should be completely redesigned to show the actual calibration steps more clearly.
4. Figure 2 should include the main processing steps needed for the calibration, including collocation, pixel screening, cloud phase selection, parallax/geolocation treatment, spectral response treatment, LUT interpolation, radiative transfer assumptions, and uncertainty propagation. At present, the figure gives only a high-level workflow and does not provide enough information for readers to understand the calibration process.
5. Figure 3 is intended for aerosol calibration, but the basic idea is similar to the cloud calibration workflow. The figure is not very informative and overlaps conceptually with Figures 1 and 2. If the authors keep it, they should make it much more specific to the aerosol case by showing the aerosol input variables, quality screening, STAR simulation, ocean-surface treatment, and coefficient-estimation procedure. Otherwise, it could be merged with Figure 2 into one clearer workflow figure.
6. The purpose of Figures 4 and 5 is unclear. Figures 4 and 5 show MSI_CLP cloud properties for selected frames, but the scientific purpose of including them is not well explained. Are these figures intended to demonstrate scene selection, cloud-property retrieval quality, or calibration-target representativeness? If they are only used to show that cloud products exist, they are not necessary. If the authors want to compare MSI and Himawari retrievals, then the corresponding MSI and Himawari products should be shown side by side using the same spatial domain, color scale, and screening mask.
7. The text discusses frame 01808D as a reference case, but the current figure set does not present it consistently with the other selected frames.
8. Figures 6–8 also need stronger scientific purpose. The Himawari-9 cloud product figures are difficult to interpret and are not directly compared with the MSI retrievals. If the authors want to use Himawari as the reference/intermediary, then the manuscript should show the matched MSI and Himawari scenes together, after applying the same mask and collocation procedure. As currently presented, these figures do not allow the reader to evaluate the quality of the matching or the representativeness of the selected pixels.
9. The color bars in Figures 9–11 are not ideal for these scatter plots because the point density and scatter width change with radiance. A better approach would be to use transparent points, a log-density hexbin plot, or a 2-D histogram with clearly defined binning.
10. Figure 12 is too small and the red box is not sufficiently explained. The red box should be explicitly described in both the text and the caption. The figure itself is too small to allow readers to understand the selected calibration area. If the red box defines the region used for Figures 14 and 15, that connection should be stated clearly.
11. The color-bar ranges in Figure 13 should be reconsidered so that spatial variations in AOT, AE, and SSA can actually be seen. The figure caption should also explain whether these are raw Himawari aerosol products, quality-screened products, or only the subset used for the MSI simulation.
12. In Figure 14, most pixels appear to be near the upper end of the color scale, making the spatial variation difficult to see. If almost no pixels are below approximately 0.03, the lower color-bar limit should not start at 0.02. The authors should choose a color range that reveals the actual variation in the selected scene. The observed and simulated panels should also use exactly the same color scale.
13. Many figures are not discussed properly in the main text. The authors should not simply cite a figure and then state a result. For each figure, they should explain what the figure is intended to demonstrate, what processing was applied, what the key quantitative result is, and how that result supports the calibration conclusion.
14. It is better to have an uncertainty discussion. A vicarious calibration study should include an uncertainty budget. For example, discuss the uncertainty contributions from Himawari calibration, Himawari cloud/aerosol retrieval errors, spectral response differences, collocation mismatch, cloud evolution, parallax, LUT interpolation, radiative transfer assumptions, surface/ocean treatment, and scene-selection bias. Without this, the calibration coefficients are difficult to interpret.
15. The manuscript contains multiple grammar errors and awkward phrases, including “lever 2 aerosol product,” “The utilization of vicarious calibration can be needed,” “which supposed to have overestimation,” and “This result has delighted both us and the ESA team.” The last phrase is not appropriate for a scientific paper and should be replaced with neutral technical wording.
16. Please clearly distinguish between radiance and reflectance throughout the manuscript.
17. Please clarify whether the calibration coefficient is intended to be global, scene-dependent, band-dependent, or product-specific. And discuss whether the same coefficient should apply across the full MSI scan, considering possible spectral response and detector-dependent behavior.