Uncertainty characterization of HOAPS 3.3 latent heat-flux-related parameters

Liman, Julian; Schröder, Marc; Fennig, Karsten; Andersson, Axel; Hollmann, Rainer

doi:https://doi.org/10.5194/amt-11-1793-2018

Articles | Volume 11, issue 3

https://doi.org/10.5194/amt-11-1793-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 3.0 License.

https://doi.org/10.5194/amt-11-1793-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 3.0 License.

Articles | Volume 11, issue 3

Research article

|

29 Mar 2018

Research article |

| 29 Mar 2018

Uncertainty characterization of HOAPS 3.3 latent heat-flux-related parameters

Julian Liman, Marc Schröder, Karsten Fennig, Axel Andersson, and Rainer Hollmann

Download

Final revised paper (published on 29 Mar 2018)
Preprint (discussion started on 31 Jul 2017)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Uncertainty Characterization of HOAPS-3.3 Latent Heat Flux Related Parameters', Anonymous Referee #1, 08 Sep 2017
- AC1: 'Answers to comments by referee #1', Julian Kinzel, 01 Dec 2017
- AC4: '(Changes implemented into the) revised manuscript', Julian Kinzel, 01 Dec 2017
RC2: 'Review', Anonymous Referee #2, 26 Oct 2017
- AC2: 'Answers to comments by referee #2', Julian Kinzel, 01 Dec 2017
- AC5: '(Changes implemented into the) revised manuscript', Julian Kinzel, 01 Dec 2017
RC3: 'AMT-2017-176 "Uncertainty characterization of HOAPS-3.3 latent heat flux related parameters", by J. Kinzel, M. Schroder, K. Fennig, A. Andersson and R. Hollmann', Anonymous Referee #3, 08 Nov 2017
- AC3: 'Answers to comments by referee #3', Julian Kinzel, 01 Dec 2017
- AC6: '(Changes implemented into the) revised manuscript', Julian Kinzel, 01 Dec 2017

Peer-review completion

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

AR by Julian Liman on behalf of the Authors (01 Dec 2017) Author's response Manuscript

ED: Referee Nomination & Report Request started (23 Dec 2017) by Ad Stoffelen

RR by Anonymous Referee #3 (02 Jan 2018)

Suggestions for revision or reasons for rejection

I read the revised paper for second review, including the comprehensive response to reviewer comments. Although the changes to the manuscript are non-negligible, I feel that the response to reviewers is more comprehensive than the revision of the manuscript. Typically, reviewers ask questions because the paper is not very clear so clarification of the manuscript is needed and it is not sufficient to reply to the reviewers.

I repeat my general comment. The work is very interesting and well worth publishing, but unfortunately the methodology is not well explained. Abstract and introduction can be understood once the paper is understood, but not the other way around. I feel that abstract and introduction should be self-explanatory.

Although I can live with this manuscript as is, to improve accessibility of the paper, I suggest to rewrite the methodology part of the abstract (say page 1, lines 4 to 10), the last paragraph of the introduction (page 4, lines 19 to 30), and the introduction of the methodology section (page 8, lines 12 to 20) from scratch.

A few suggestions:
1.
The lack of clarity comes from the use of jargon that is clear for the authors but not for the reader. Jargon like "double collocation analysis", "multi-dimensional bias analysis", and "random uncertainty decomposition" may not mean much to readers that are not familiar with your work. It would be good to explain the basic ideas of this work in abstract, introduction and methodology sections. The paper is in my view based on the following ideas: (i) systematic errors can be estimated by assuming that in-situ observations are bias free, (ii) random errors can be estimated by considering three different data sources at the same location (called triple collocation), and (iii) the error estimates can be generalized from locations where triple co-location is possible to global coverage, by assuming that the systematic and random errors are uniquely related to predictors that can be observed by satellite (namely, q_a, U, SST, wvpa). I suggest to elaborate on these 3 points to describe clearly what this paper is about.

2.
The reference to Kinzel et al. (2016) on page 4, line 21 is still rather obscure as part of the introduction. In the introduction I suggest to explain: what is covered in the Kinzel et al. paper and what is new in the current work.

3.
Page 9 lines 17 to 21 addresses the issue of correlated variables on absis and ordinate, which is good. I guess that on line 18, q_a, U and SST are from in situ observations rather than satellite, but it is not mentioned explicitly. I am confused, because the same symbols are used as on line 2.

Hide

RR by Anonymous Referee #1 (06 Jan 2018)

ED: Publish subject to minor revisions (review by editor) (14 Jan 2018) by Ad Stoffelen

AR by Julian Liman on behalf of the Authors (22 Jan 2018) Author's response Manuscript

ED: Publish subject to minor revisions (review by editor) (15 Feb 2018) by Ad Stoffelen

The authors have clearly made the manuscript more accessible to the readers after addressing the suggestions by the reviewers, which is laudable. Nevertheless, there remain several main instances where the paper provides a poor insight and I'd appreciate if the manuscript was further clarified on these topics before publication:
- The methodology appears to rely on conditional sampling, e.g., if x and y measure truth t, then the mean of y-x for given x is evaluated and interpreted as the bias (or SD) of y versus t. This is incorrect of x has an error. The error of x will broaden the distribution of x w.r.t. t and therefore extreme x are larger than extreme t, such that y should be biased against x, even though x and y have only random error w.r.t. t (See, e.g., Stoffelen, 1998); See, inter alia, Figure 2.
- It is assumed that errors in q, U and T are independent, which appears rather odd? In areas with high SST variability, high wind variability will occur, as well as variability in humidity and air temperature. Also, in moist convection, all atmospheric parameters tend to vary simultaneously and in correlated ways. Besides correlations expected from physical processes, correlations are also expected in the simultaneous retrievals. A retrieval is an algorithm where radiance measurements are compromised in order to obtain a geophysical retrieval. Ergo, an error in U in the retrieval is likely associated with compensating errors in the other retrieved geophysical variables. Furthermore, if the same QC, cal/val and retrieval algorithms are used for two different instruments, error correlation of those multi-variable retrievals is likely too. This should be made clear.
- The error model is not clear formally. What is assumed to be truth and what instrumental and geophysical variations are exactly captured by the error variables?
- The manuscript does not appear to clearly separate statistical and geophysical effects. Of course, statistical results may be linked to physical processes, but also to coincidental occurrence of high humidity and low winds, for example of conditional binning artifacts.
- Assumptions made on sampling, correlations, error models andsoforth have ever-lasting effects on the results obtained, even after a trillion collocations have been evaluated.
I trust that a further appraisal of the above critical points, while going through the manuscript, will provide further clarity to your potential AMT readership.

Hide

AR by Marc Schroeder on behalf of the Authors (23 Feb 2018) Author's response Manuscript

ED: Publish as is (23 Feb 2018) by Ad Stoffelen

AR by Marc Schroeder on behalf of the Authors (26 Feb 2018)

Short summary

Latent heat fluxes (LHF) play a major role in the climate system. Over open ocean, they are increasingly observed by satellite instruments. To access their quality, this research focuses on thorough uncertainty analysis of all LHF-related variables of the HOAPS satellite climatology, in parts making use of novel analysis approaches. Results indicate climatological LHF uncertainies up to 50 W m⁻², whereby underlying specific humidities tend to be more uncertain than contributing wind speeds.