Across-track Extension of Retrieved Cloud and Aerosol Properties for the EarthCARE Mission: The ACM-3D Product

Qu, Zhipeng; Barker, Howard W.; Cole, Jason N. S.; Shephard, Mark W.

doi:https://doi.org/10.5194/amt-2022-301

Preprints

https://doi.org/10.5194/amt-2022-301

Preprints

21 Nov 2022

| 21 Nov 2022

Status: a revised version of this preprint was accepted for the journal AMT and is expected to appear here in due course.

Across-track Extension of Retrieved Cloud and Aerosol Properties for the EarthCARE Mission: The ACM-3D Product

Zhipeng Qu, Howard W. Barker, Jason N. S. Cole, and Mark W. Shephard

Abstract. The narrow cross-section of cloud and aerosol properties retrieved by L2-algorithms that operate on data from EarthCARE’s nadir-pointing sensors gets “broadened” across-track by an algorithm that is described and demonstrated here. This Scene Construction Algorithm (SCA) consists of four sub-algorithms. At its core is a radiance-matching procedure that works with measurements made by EarthCARE’s Multi-Spectral Imager (MSI). In essence, an off-nadir pixel gets filled with retrieved profiles that are associated with a (nearby) nadir pixel whose MSI radiances best match those of the off-nadir pixel. The SCA constructs a 3D array of cloud and aerosol (and surface) properties for entire frames that measure ~6,000 km along-track by 150 km across-track (i.e., the MSI’s full swath). Constructed domains out to ~15 km on both sides of nadir are used explicitly downstream as input for 3D radiative transfer models that predict top-of-atmosphere (TOA) broadband solar and thermal fluxes and radiances. These quantities are compared to com-mensurate measurements made by EarthCARE’s BroadBand Radiometer (BBR), thus facilitating a continuous closure assessment of the retrievals. Three 6,000 km x 200 km frames of synthetic EarthCARE observations were used to demonstrate the SCA. The main conclusion is that errors in modelled TOA fluxes that stem from use of 3D domains produced by the SCA are expected to be less than ±5 W m^-2 and rarely larger than ±10 W m^-2. As such, the SCA, as purveyor of in-formation needed to run 3D radiative transfer models, should help more than hinder the radiative closure assessment of EarthCARE’s L2 retrievals.

Received: 31 Oct 2022 – Discussion started: 21 Nov 2022

Zhipeng Qu et al.

Status: closed

RC1:
'Comment on amt-2022-301', Anonymous Referee #1, 19 Dec 2022

The authors describe the scene construction algorithm (SCA) for EarthCARE Level-2 data products. The algorithm is based on earlier work by Barker et al. (2011, 2012) that uses spectral radiances to transfer cloud and aerosol vertical profiles derived over ground-track of active sensors to cross-track pixels. The manuscript includes descriptions of three stages of screening process and how to determine buffer zones. The authors define the error due to SCA as the domain averaged radiance difference between observed radiances and transplanted radiances, scaled by the mean flux over the ground-track portion of the domain. The authors show that the error due to SCA is well below 10 Wm-2, which is the error budget of EarthCARE including all algorithms.

The manuscript is well written and easy to follow. I only have minor comments.

General comments

The paper does not discuss what channels/wavelengths are used for the scene construction algorithm and the total number of ks described on line 95. Channels used in the SCA are probably different for day and night. But do they also vary depending on scenes/locations? Some details are discussed in Barker et al. (2011). But it is not obvious from their paper which combination of cannels is used in the actual SCA. Could you include descriptions of channels/wavelengths used in the SCA?

One of conclusions is that the error by the SCA is less than 5 Wm-2 or 3 Wm-2. However, this is based on results using the Hawaii frame (line 276 to 280). Is this also true for other two frames used for testing?

Specific comments

Abstract

It is not described anywhere in the manuscript what four sub-algorithms are.

The last sentence of the abstract and Section 3.3.

The direct way to show that the SCA helps more than hinder toward achieving the radiative closure goal of EarthCARE is to show that TOA flux error with and without the SCA. But this study did not address that. The SCA can help reducing the TOA flux error in two says. One way is by identifying clear-sky for the entire BBR footprint. If I look Figure 2 of Ham et al. (2015), nearly 30% of along-track clear-sky scenes contains up to 10% clouds. The SCA should reduce the TOA flux error identifying clouds present off-nadir. Second, the SCA can provide better off-nadir cloud information than no information. Top two plots of Figure 4 of Ham et al. (2015) show improvements of TOA fluxes (smaller differences between CERES-derived and computed fluxes) with the SCA. Also, the left plot of Figure 6 shows the improvement for almost all cloud types. If the authors prefer to estimate in their way, both effects together can be estimated by limiting the area of averaging radiance just over along-track in Eq. (5).

Section 2.3

The authors describe three stages of screen processes. If a domain contains corrupted data are rejected (line 154), I am wondering what is the fraction of domains that pass this screen process. Do you have an estimate of how often domains are rejected? Could you include the number (yield) based on scenes the authors worked on so far? If active sensor retrievals systematically fail for certain type of clouds, such as deep convective clouds, then these clouds have never been included in the radiative closure assessment. Could you include authors thoughts/concerns that the closure is performed preferably toward certain cloud types?

One of variables used for screening is the solar zenith angle. Currently, the threshold of the solar zenith angle is 75 degrees. In addition, homogeneous land surface and standard deviation of surface elevations are used to screen scenes to be used for radiative closure. Can the authors estimate the faction of Ds that pass these screening process?

Line 213. Could you explain what 2020 mean?

Reference

Ham, S.H., S. Kato, H. W. Barker, F. G. Rose, and S. Sun-Mack, Improving the modeling of short-wave radiation through the use of a 3D scene construction algorithm, Q. J. R. Meteorol. Soc., (2015), DOI: 10.1002/qj.2491.

Citation: https://doi.org/10.5194/amt-2022-301-RC1
- AC1: 'Reply on RC1', Zhipeng Qu, 09 Feb 2023
  
  The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-301/amt-2022-301-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/amt-2022-301-AC1
RC2:
'Comment on amt-2022-301', Anonymous Referee #2, 10 Jan 2023
General comments

The authors present a simple yet elegant concept for building full three-dimensional atmospheric profiles. The exposition is detailed but very clear (with a few minor comments that I will address shortly). The benefits to the EarthCARE mission are obvious and immediate, yet the authors are not blind to the limitations of the proposed approach (just to give one example, the discussion of Figs. 7 and 9 contains a fair appraisal of the algorithm's estimated error). It is clear that this manuscript does not constitute a full demonstration of the algorithm's performance, but I also appreciate that this was not possible within the scope of this submission, and the authors do include a reference to another contribution (that I have not reviewed). Technically, the paper appears sound, even though I do not feel fully qualified to review all technical details (especially appendix B).

I would rate the manuscript length as being perhaps on the longish side, but without being excessive.

Specific comments

The term "assessment domain" (line 54) and the processor designations ACM-RT and ACMB-DF (lines 74 and 75) have not been formally introducted before first use. However, this may be resolved at the time of publication (I did not check the referenced publications).

What is not immediately clear in the main body of the text, is whether the SCA works with one channel at a time, or with a combination of channels. Although this is addressed in the appendix (it is able to use any conceivable, weighted combination), I would suggest explicitly adding this to the main text to improve linear reading, as I believe this influences the interpretation of the results. (Otherwise, I cannot explain the increased error when moving away from the ground track.)

Formula (1) uses a minimum, whereas I would have naïvely expected a maximum over the domain, to ensure all values remain bounded by above. Could the authors explain this?

The discussion in 3.1 mention the increasing RMSE with increasing across-track distance from the ground track (visible in Fig. 7). My interpretation is that, with increasing across-track distance, meteorological conditions start to differ more, and it becomes increasingly difficult to match the ensemble of MSI radiances with satisfying accuracy. I think the text would benefit from a brief explanation by the authors.

The role of Lambda in Appendix A (line 345) appears to be a selector for suitable donor/acceptor pairs. I take it that any such pairs with F exactly equal to zero disqualify immediately for the selection in formula (15), because they would otherwise falsify the ranking? This is not stated explicitly in the text.

Appendix B felt quite technical, and I admit that I could not fully grasp the technical details.

Figure 5 does not mention or show m_buffer (across-track) or clouds, is this the figure that was intended to be shown?

In Figure 6, I would suggest adding a marginal (side) plot to the left of the left half of channel 1, plotting channel 1 nadir radiance as a function of latitude (and similarly for channel 4), to assist the reader in assessing the working of the algorithm. For the case shown, I would expect a relatively flat and relatively low nadir radiance profile between ~10 and 13 degrees N, which would explain the gradual failure of the algorithm in matching the high off-nadir radiances, and the visible "banding" in the reconstructed radiances.

Suggested technical corrections

What follows is a list of suggestions that I would humbly propose. Please note that I am not a native speaker, so I respectfully defer to the editor and authors for any final decision.

14 (abstract) "out to ~15 km on both sides of nadir" : suggest adding "along-track" explicitly for clarity

48-49 possible grammar mistake in "The extreme case is use"

66 typo in "retreived"

83 possible missing article in "At its core is definition of D"

114-115 possible redundancy in "adverse effects near the perimeter of D are affected by [...]"

119 possible grammar mistake in "Buffer-zone also accomodate"

130 semicolon where a comma was expected

133 semicolon where a comma was expected

195 symbol (µ) appears to be missing in wavelength units (twice)

213 possible typo in "where"

214 suggest adding a comma after "and for each array"

254 text unclear in "[values] get reduced by at least a factor of 10 when for averages [...]"

309-310 a verb appears to be missing in the subsentence spanning these 2 lines

317 possible typo in "cost-effect"
Citation: https://doi.org/10.5194/amt-2022-301-RC2
- AC2: 'Reply on RC2', Zhipeng Qu, 09 Feb 2023
  
  The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-301/amt-2022-301-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/amt-2022-301-AC2

Status: closed

RC1:
'Comment on amt-2022-301', Anonymous Referee #1, 19 Dec 2022

The authors describe the scene construction algorithm (SCA) for EarthCARE Level-2 data products. The algorithm is based on earlier work by Barker et al. (2011, 2012) that uses spectral radiances to transfer cloud and aerosol vertical profiles derived over ground-track of active sensors to cross-track pixels. The manuscript includes descriptions of three stages of screening process and how to determine buffer zones. The authors define the error due to SCA as the domain averaged radiance difference between observed radiances and transplanted radiances, scaled by the mean flux over the ground-track portion of the domain. The authors show that the error due to SCA is well below 10 Wm-2, which is the error budget of EarthCARE including all algorithms.

The manuscript is well written and easy to follow. I only have minor comments.

General comments

The paper does not discuss what channels/wavelengths are used for the scene construction algorithm and the total number of ks described on line 95. Channels used in the SCA are probably different for day and night. But do they also vary depending on scenes/locations? Some details are discussed in Barker et al. (2011). But it is not obvious from their paper which combination of cannels is used in the actual SCA. Could you include descriptions of channels/wavelengths used in the SCA?

One of conclusions is that the error by the SCA is less than 5 Wm-2 or 3 Wm-2. However, this is based on results using the Hawaii frame (line 276 to 280). Is this also true for other two frames used for testing?

Specific comments

Abstract

It is not described anywhere in the manuscript what four sub-algorithms are.

The last sentence of the abstract and Section 3.3.

The direct way to show that the SCA helps more than hinder toward achieving the radiative closure goal of EarthCARE is to show that TOA flux error with and without the SCA. But this study did not address that. The SCA can help reducing the TOA flux error in two says. One way is by identifying clear-sky for the entire BBR footprint. If I look Figure 2 of Ham et al. (2015), nearly 30% of along-track clear-sky scenes contains up to 10% clouds. The SCA should reduce the TOA flux error identifying clouds present off-nadir. Second, the SCA can provide better off-nadir cloud information than no information. Top two plots of Figure 4 of Ham et al. (2015) show improvements of TOA fluxes (smaller differences between CERES-derived and computed fluxes) with the SCA. Also, the left plot of Figure 6 shows the improvement for almost all cloud types. If the authors prefer to estimate in their way, both effects together can be estimated by limiting the area of averaging radiance just over along-track in Eq. (5).

Section 2.3

The authors describe three stages of screen processes. If a domain contains corrupted data are rejected (line 154), I am wondering what is the fraction of domains that pass this screen process. Do you have an estimate of how often domains are rejected? Could you include the number (yield) based on scenes the authors worked on so far? If active sensor retrievals systematically fail for certain type of clouds, such as deep convective clouds, then these clouds have never been included in the radiative closure assessment. Could you include authors thoughts/concerns that the closure is performed preferably toward certain cloud types?

One of variables used for screening is the solar zenith angle. Currently, the threshold of the solar zenith angle is 75 degrees. In addition, homogeneous land surface and standard deviation of surface elevations are used to screen scenes to be used for radiative closure. Can the authors estimate the faction of Ds that pass these screening process?

Line 213. Could you explain what 2020 mean?

Reference

Ham, S.H., S. Kato, H. W. Barker, F. G. Rose, and S. Sun-Mack, Improving the modeling of short-wave radiation through the use of a 3D scene construction algorithm, Q. J. R. Meteorol. Soc., (2015), DOI: 10.1002/qj.2491.

Citation: https://doi.org/10.5194/amt-2022-301-RC1
- AC1: 'Reply on RC1', Zhipeng Qu, 09 Feb 2023
  
  The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-301/amt-2022-301-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/amt-2022-301-AC1
RC2:
'Comment on amt-2022-301', Anonymous Referee #2, 10 Jan 2023
General comments

The authors present a simple yet elegant concept for building full three-dimensional atmospheric profiles. The exposition is detailed but very clear (with a few minor comments that I will address shortly). The benefits to the EarthCARE mission are obvious and immediate, yet the authors are not blind to the limitations of the proposed approach (just to give one example, the discussion of Figs. 7 and 9 contains a fair appraisal of the algorithm's estimated error). It is clear that this manuscript does not constitute a full demonstration of the algorithm's performance, but I also appreciate that this was not possible within the scope of this submission, and the authors do include a reference to another contribution (that I have not reviewed). Technically, the paper appears sound, even though I do not feel fully qualified to review all technical details (especially appendix B).

I would rate the manuscript length as being perhaps on the longish side, but without being excessive.

Specific comments

The term "assessment domain" (line 54) and the processor designations ACM-RT and ACMB-DF (lines 74 and 75) have not been formally introducted before first use. However, this may be resolved at the time of publication (I did not check the referenced publications).

What is not immediately clear in the main body of the text, is whether the SCA works with one channel at a time, or with a combination of channels. Although this is addressed in the appendix (it is able to use any conceivable, weighted combination), I would suggest explicitly adding this to the main text to improve linear reading, as I believe this influences the interpretation of the results. (Otherwise, I cannot explain the increased error when moving away from the ground track.)

Formula (1) uses a minimum, whereas I would have naïvely expected a maximum over the domain, to ensure all values remain bounded by above. Could the authors explain this?

The discussion in 3.1 mention the increasing RMSE with increasing across-track distance from the ground track (visible in Fig. 7). My interpretation is that, with increasing across-track distance, meteorological conditions start to differ more, and it becomes increasingly difficult to match the ensemble of MSI radiances with satisfying accuracy. I think the text would benefit from a brief explanation by the authors.

The role of Lambda in Appendix A (line 345) appears to be a selector for suitable donor/acceptor pairs. I take it that any such pairs with F exactly equal to zero disqualify immediately for the selection in formula (15), because they would otherwise falsify the ranking? This is not stated explicitly in the text.

Appendix B felt quite technical, and I admit that I could not fully grasp the technical details.

Figure 5 does not mention or show m_buffer (across-track) or clouds, is this the figure that was intended to be shown?

In Figure 6, I would suggest adding a marginal (side) plot to the left of the left half of channel 1, plotting channel 1 nadir radiance as a function of latitude (and similarly for channel 4), to assist the reader in assessing the working of the algorithm. For the case shown, I would expect a relatively flat and relatively low nadir radiance profile between ~10 and 13 degrees N, which would explain the gradual failure of the algorithm in matching the high off-nadir radiances, and the visible "banding" in the reconstructed radiances.

Suggested technical corrections

What follows is a list of suggestions that I would humbly propose. Please note that I am not a native speaker, so I respectfully defer to the editor and authors for any final decision.

14 (abstract) "out to ~15 km on both sides of nadir" : suggest adding "along-track" explicitly for clarity

48-49 possible grammar mistake in "The extreme case is use"

66 typo in "retreived"

83 possible missing article in "At its core is definition of D"

114-115 possible redundancy in "adverse effects near the perimeter of D are affected by [...]"

119 possible grammar mistake in "Buffer-zone also accomodate"

130 semicolon where a comma was expected

133 semicolon where a comma was expected

195 symbol (µ) appears to be missing in wavelength units (twice)

213 possible typo in "where"

214 suggest adding a comma after "and for each array"

254 text unclear in "[values] get reduced by at least a factor of 10 when for averages [...]"

309-310 a verb appears to be missing in the subsentence spanning these 2 lines

317 possible typo in "cost-effect"
Citation: https://doi.org/10.5194/amt-2022-301-RC2
- AC2: 'Reply on RC2', Zhipeng Qu, 09 Feb 2023
  
  The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-301/amt-2022-301-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/amt-2022-301-AC2

Zhipeng Qu et al.

Viewed

Total article views: 350 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
250	85	15	350	7	6

HTML: 250
PDF: 85
XML: 15
Total: 350
BibTeX: 7
EndNote: 6

Views and downloads (calculated since 21 Nov 2022)

Month	HTML	PDF	XML	Total
Nov 2022	98	31	5	134
Dec 2022	48	18	3	69
Jan 2023	45	10	2	57
Feb 2023	44	18	5	67
Mar 2023	15	8	0	23

Cumulative views and downloads (calculated since 21 Nov 2022)

Month	HTML	PDF	XML	Total
Nov 2022	98	31	5	134
Dec 2022	48	18	3	69
Jan 2023	45	10	2	57
Feb 2023	44	18	5	67
Mar 2023	15	8	0	23

Viewed (geographical distribution)

Total article views: 361 (including HTML, PDF, and XML) Thereof 361 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 25 Mar 2023

Short summary

This paper describes EarthCARE’s L2 product ACM-3D. It includes the Scene Construction Algorithm (SCA) used to produce the indexes for reconstructing 3D atmospheric scene based on satellite nadir retrievals. It also provide the information about the buffer zone sizes of 3D assessment domains and the ranking scores for selecting the best 3D assessment domains. These output variables are needed to run 3D radiative transfer models for the radiative closure assessment of EarthCARE’s L2 retrievals.


Total:	0
HTML:	0
PDF:	0
XML:	0