The Community Cloud retrieval for CLimate (CC4CL) – Part 1: A framework applied to multiple satellite imaging sensors

Sus, Oliver; Stengel, Martin; Stapelberg, Stefan; McGarragh, Gregory; Poulsen, Caroline; Povey, Adam C.; Schlundt, Cornelia; Thomas, Gareth; Christensen, Matthew; Proud, Simon; Jerg, Matthias; Grainger, Roy; Hollmann, Rainer

doi:https://doi.org/10.5194/amt-11-3373-2018

Articles | Volume 11, issue 6

https://doi.org/10.5194/amt-11-3373-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/amt-11-3373-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 11, issue 6

Research article

|

13 Jun 2018

Research article |

| 13 Jun 2018

The Community Cloud retrieval for CLimate (CC4CL) – Part 1: A framework applied to multiple satellite imaging sensors

Oliver Sus, Martin Stengel, Stefan Stapelberg, Gregory McGarragh, Caroline Poulsen, Adam C. Povey, Cornelia Schlundt, Gareth Thomas, Matthew Christensen, Simon Proud, Matthias Jerg, Roy Grainger, and Rainer Hollmann

Download

Final revised paper (published on 13 Jun 2018)
Preprint (discussion started on 17 Oct 2017)

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'Reviewer Comment', Anonymous Referee #2, 15 Nov 2017
- AC3: 'Authors responses to referee #2', Oliver Sus, 05 Mar 2018
RC2: 'Review of Amt-2017-334', Anonymous Referee #1, 31 Dec 2017
- AC1: 'Authors responses to referee #1', Oliver Sus, 05 Mar 2018
RC3: 'Reviewer Comment', Anonymous Referee #3, 08 Jan 2018
- AC2: 'Authors responses to referee #3', Oliver Sus, 05 Mar 2018

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by Oliver Sus on behalf of the Authors (09 Apr 2018) Author's response Manuscript

ED: Referee Nomination & Report Request started (23 Apr 2018) by Brian Kahn

RR by Anonymous Referee #3 (04 May 2018)

RR by Anonymous Referee #1 (04 May 2018)

Suggestions for revision or reasons for rejection

The general impression from the response is that the authors are pretty much set on their approach. The use of ANNCOD in the background without quantifying its quality relative to CALIPSO is unfortunate, and it casts some doubt on the manuscript. I do think that a comparison of ANNCOD with CALIPSO makes sense, at least for COD<3. If this is what is used for the cloud mask, then it should be quality assessed. For bright regions, the authors' statement that passive sensors will always detect clouds for COD>1 (in their response) is incorrect. At some point, the authors comment that ANNCOD is a pseudo optical depth, which is "relative". But if that is the case, a formal definition should be provided what "pseudo" means. Following the current manuscript, ANNCOD seems to be unscaled, and trained with CALIPSO data, so it should reproduce CALIPSO data. However, the authors state that "ANNCOD does not provide any information on the absolute optical thickness value." At that point one has to wonder what it does do. If ANNCOD does not contain any information on COD, then the name "ANNCOD" is misleading, and the product should not be used as the basis for the cloud mask.

Otherwise, the added detail with regard to NN is helpful. However, questions remain how well it works with one single hidden layer. Just because the used IDL library does not provide more than one layer is not a good justification for limiting oneself to something that may not be adequate for the problem. Might this shallow NN be the reason that the authors don't show the validation against CALIPSO COD? In a similar vein, it is unfortunate that the authors do not want to show the histograms from figure 7 for snow bright surfaces (their response to the reviewer's comment to figure 7).

A few minor comments:
p2,l49: "insufficient knowledge...propagates uncertainties..." still doesn't sound quite right. It is not the in sufficient knowledge that propagates uncertainties. Just a language issue.

p2,l78: "Consistency in approach can be traded for continuity of results, and multi-platform algorithms could exploit additional data when newer sensors become available" is still unclear and needs elaboration.

"For a single layer, optically thick (COT>1)" -- This is incorrect. Cloud with COD>1 are not optically thick.

Regarding the t-test (p13,l39): It is understood that readers will be familiar with a t-test, that was not the point of the comment.

p13,l45: "spatiotemporally collocated sensors" The comment was not about the mathematical meaning of the intercomparison, but about the physical meaning. Why *are* the data different, even though they are collocated?

Hide

RR by Anonymous Referee #2 (09 May 2018)

Suggestions for revision or reasons for rejection

p 2, lines 34-35: I think the authors inserted the Marchant reference and description in the wrong sentence. That paper discusses the MODIS C6 optical property phase, not cloud detection/masking, and fits into the next sentence. Moreover, the next sentence discussing phase should include the Baum et al (2012) reference also since it is referring to the MODIS C6 IR phase which uses 8.5 and 11µm in addition to 7.3 and 12µm (the authors provide an incorrect description of this). There are two independent phase algorithms in MODIS C6 and both should be referenced.

p 3, lines 6-7: “and beyond a penetration depth into the cloud corresponding to > 1 cumulative optical depth (Baum et al., 2012).” I’m not sure how this in itself describes an insensitivity to cloud top properties. The roughly 1 optical depth penetration depth is simply the reason why the MODIS C6 cloud top retrieval (and any other IR cloud top retrieval) loses sensitivity for optically thin clouds and why it cannot be considered the physical cloud top.

p 3, line 22: “However, the detection of optically thin ice clouds over warm, bright surfaces remains problematic (Marchant et al., 2016).” This should actually refer to the phase identification of optically thin ice clouds, which is what the Marchant paper discusses, not the detection, which is a cloud mask issue.

p 5, line 21: CALIOP is the lidar itself, CALIPSO is the satellite platform.

p 6, line 8: “on-board” instead of “on-bard”

p 16, lines 15-16: “In particular it would be worth investigating the impact of spectral response differences, which was outside the scope of this paper and the ESA Cloud_cci project.” Adding this statement is fine, but it is now at odds with previous statements (e.g., same page lines 1-2) that spectral response differences are unlikely the causes of retrieval differences, and with the immediately following statement.

Hide

ED: Publish subject to minor revisions (review by editor) (21 May 2018) by Brian Kahn

AR by Oliver Sus on behalf of the Authors (30 May 2018) Author's response Manuscript

ED: Publish as is (05 Jun 2018) by Brian Kahn

AR by Oliver Sus on behalf of the Authors (05 Jun 2018)

Short summary

This paper presents a new cloud detection and classification framework, CC4CL. It applies a sophisticated optimal estimation method to derive cloud variables from satellite data of various polar-orbiting platforms and sensors (AVHRR, MODIS, AATSR). CC4CL provides explicit uncertainty quantification and long-term consistency for decadal timeseries at various spatial resolutions. We analysed 5 case studies to show that cloud height estimates are very realistic unless optically thin clouds overlap.