Data reduction for inverse modeling: an adaptive approach v1.0

Liu, Xiaoling; Weinbren, August L.; Chang, He; Tadić, Jovan M.; Mountain, Marikate E.; Trudeau, Michael E.; Andrews, Arlyn E.; Chen, Zichong; Miller, Scot M.

doi:https://doi.org/10.5194/gmd-14-4683-2021

Articles | Volume 14, issue 7

https://doi.org/10.5194/gmd-14-4683-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/gmd-14-4683-2021

© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 14, issue 7

Development and technical paper

|

29 Jul 2021

Development and technical paper |

| 29 Jul 2021

Data reduction for inverse modeling: an adaptive approach v1.0

Xiaoling Liu, August L. Weinbren, He Chang, Jovan M. Tadić, Marikate E. Mountain, Michael E. Trudeau, Arlyn E. Andrews, Zichong Chen, and Scot M. Miller

Download

Final revised paper (published on 29 Jul 2021)
Supplement to the final revised paper
Preprint (discussion started on 30 Sep 2020)
Supplement to the preprint

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Review for paper', Anonymous Referee #1, 12 Jan 2021
RC2: 'review comments', Anonymous Referee #2, 12 Jan 2021
AC1: 'Replies to the reviewers', Scot Miller, 30 Mar 2021

Peer-review completion

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

AR by Scot Miller on behalf of the Authors (01 Apr 2021) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (17 Apr 2021) by Havala Pye

RR by Anonymous Referee #2 (06 May 2021)

Suggestions for revision or reasons for rejection

It’s still not clear to me whether or not the authors have correctly applied the observation operator in their inverse modeling. Here in this study, the observation operator should include application of satellite a priori, averaging kernels, pressure weighting function from OCO-2 level2 data files. Please refer to Cogan et al. (2012) for the application of observation operator in comparison between model and satellite XCO2 observations. The application of averaging kernel is very important when comparing model and satellite observations, as this can reduce the impact from satellite a priori. In your synthetic cases, you can also generate synthetic satellite data using model simulation and satellite a priori, averaging kernels, pressure weighting function from OCO-2 level2 data files. Correctly using observation operator or not can impact your top-down CO2 fluxes from both full-data inversion and reduced-data inversion. Thus this can impact the evaluation of your data reduction algorithm.
I recommend publication after the authors clarifying this part and addressing the following questions/comments.

Specific comments:
Line 52-53: “Rather, these models are often used to calculate the product of H or HT and a vector (e.g., a vector of estimated CO2 fluxes)”. Here, a vector of estimated CO2 fluxes is not an appropriate example. Usually, the vector is the gradients of your objective function with respect to simulated targeted species, here in your study, simulated CO2 profiles.
Line 55-56: “The model output must be interpolated to the locations of the observations, often referred to as the observation operator.” Partially correct. The key part of the observation operator, here in your study, is the application of satellite a priori, averaging kernels and pressure weighting function.
Line 59-66: I wonder how different are the gradients at the first iteration between using reduced data and using full data, in terms of spatial distribution and magnitude?

Reference:
Cogan, A. J., et al. (2012), Atmospheric carbon dioxide retrieved from the Greenhouse gases Observing SATellite (GOSAT): Comparison with ground‐based TCCON observations and GEOS‐Chem model calculations, J. Geophys. Res., 117, D21301, doi:10.1029/2012JD018087.

Hide

ED: Publish subject to minor revisions (review by editor) (11 May 2021) by Havala Pye

AR by Scot Miller on behalf of the Authors (02 Jun 2021) Author's response Author's tracked changes Manuscript

ED: Publish as is (10 Jun 2021) by Havala Pye

Short summary

Observations of greenhouse gases have become far more numerous in recent years due to new satellite observations. The sheer size of these datasets makes it challenging to incorporate these data into statistical models and use these data to estimate greenhouse gas sources and sinks. In this paper, we develop an approach to reduce the size of these datasets while preserving the most information possible. We subsequently test this approach using satellite observations of carbon dioxide.