Technical Note: Atmospheric CO<sub>2</sub> inversions on the mesoscale using data-driven prior uncertainties: methodology and system evaluation

Kountouris, Panagiotis; Gerbig, Christoph; Rödenbeck, Christian; Karstens, Ute; Koch, Thomas Frank; Heimann, Martin

doi:https://doi.org/10.5194/acp-18-3027-2018

Articles | Volume 18, issue 4

https://doi.org/10.5194/acp-18-3027-2018

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

Special issue:

Data assimilation in carbon/biogeochemical cycles: consistent...

https://doi.org/10.5194/acp-18-3027-2018

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

Articles | Volume 18, issue 4

Technical note

|

02 Mar 2018

Technical note |

| 02 Mar 2018

Technical Note: Atmospheric CO₂ inversions on the mesoscale using data-driven prior uncertainties: methodology and system evaluation

Panagiotis Kountouris, Christoph Gerbig, Christian Rödenbeck, Ute Karstens, Thomas Frank Koch, and Martin Heimann

Download

Final revised paper (published on 02 Mar 2018)
Preprint (discussion started on 20 Sep 2016)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Review of Kountouris et al., 2016', Anonymous Referee #3, 25 Oct 2016
RC2: 'Kountouris et al. “Atmospheric CO2 inversions at the mesoscale using data driven prior uncertainties. Part 1: Methodology and system evaluation”.', Anonymous Referee #2, 25 Oct 2016
RC3: 'Review of “Atmospheric CO2 inversions at the mesoscale using data driven prior uncertainties. Part 1: Methodology and system evaluation” Kountouris et al., 2016', Anonymous Referee #1, 11 Nov 2016
AC1: 'Author's response', Panagiotis Kountouris, 30 Jan 2017

Peer-review completion

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

AR by Panagiotis Kountouris on behalf of the Authors (01 Mar 2017) Author's response

ED: Referee Nomination & Report Request started (14 Mar 2017) by Yafang Cheng

RR by Anonymous Referee #3 (01 Apr 2017)

Suggestions for revision or reasons for rejection

The revised paper has clarified some material, compared to the earlier version. However, the changes are relatively minor, and have not allayed my concerns that the material is too incremental to warrant a stand-alone publication. In particular:

1. My previous comment still stands that improvements in "goodness of fit" tests (RMSE, correlation, chi-squared) are simply a consequence of the inversion code performing in a way that is not obviously erroneous. In their response the authors state that an improvement in flux space is not a necessarily condition of an improvement in concentration space. Whilst this is true in the real world (in contrast to some of the comments in the revised paper, see below), it should absolutely not be the case in the situation where the model and data are perfect, as they are here. Think of the limiting case: if model-data error was very small, and prior flux error was large, one should be able to derive very close to "perfect" fluxes in the case of a perfect model and perfect data. By reducing the data constraint, one simply moves the flux solution back from "perfection" towards the prior.

2. My second general comment highlighted several papers that had developed methodology in this area in recent years. Whilst the authors have now cite a couple of them, I was hoping that, since this is a "methodology" paper, they would compare and contrast the methodology in this paper with the recent developments by other authors. However, instead they simply say that eddy covariance data are not available for the gases used in the other publications. This misses the point to me; if these papers applied their methodology to CO2 (and there is no reason why they couldn't), would the limitations identified by the authors still apply? If so, what would be the benefits of the proposed system.

Additional comments:

- In the revised paper the authors now state that no error was added to the model/data. Wouldn't it have been more interesting to include some uncertainty here and see how it influenced the mismatch between the "true" and a posteriori flux. This follows directly from comment 1 above. Without any error in the model/data, the improvement in fit is fairly trivial.

- P23 L17, states that model transport error is not "excessively assessed". This is not true. Model transport error is ignored.

- P23 L22 now states that "for coarser spatial scales transport morels perform better, and as long as the fitting performance shows good results, flux estimates should be more reliable." The first part of this sentence is an unproven assertion (that better resolution = better transport), and the second part is not correct (that better fit to the data = more reliable flux). I could have a terrible transport model that fits the data perfectly simply by compensating for transport error by deriving completely wrong fluxes.

- P25 L15 states that "Concerns might rise that the observational uncertainties are underestimated due to the absence of the error correlations. However the χr2 values prove the opposite." This may be true in this synthetic case, but it won't hold at all in the real world. In reality there must be correlated errors in chemical transport models.

- My previous comment that the comparison with pseudo-EC data still stands, and has not been addressed in the revised manuscript. The authors cannot conclude from this study that EC data can be used to validate these fluxes in the real world. In the real world, scaling issues will be absolutely critical (EC data being representative of scales that are typically an order of magnitude smaller than the transport model).

Hide

ED: Reconsider after major revisions (08 Jun 2017) by Yafang Cheng

AR by Panagiotis Kountouris on behalf of the Authors (05 Oct 2017) Author's response

ED: Publish subject to minor revisions (review by editor) (26 Oct 2017) by Yafang Cheng

AR by Panagiotis Kountouris on behalf of the Authors (31 Oct 2017) Author's response Manuscript

ED: Publish as is (06 Nov 2017) by Yafang Cheng

AR by Panagiotis Kountouris on behalf of the Authors (06 Nov 2017) Manuscript

Technical Note: Atmospheric CO2 inversions on the mesoscale using data-driven prior uncertainties: methodology and system evaluation

Download

Interactive discussion

Peer-review completion

Technical Note: Atmospheric CO₂ inversions on the mesoscale using data-driven prior uncertainties: methodology and system evaluation