Evaluation of wetland methane emissions across North America using atmospheric data and inverse modeling

Miller, Scot M.; Commane, Roisin; Melton, Joe R.; Andrews, Arlyn E.; Benmergui, Joshua; Dlugokencky, Edward J.; Janssens-Maenhout, Greet; Michalak, Anna M.; Sweeney, Colm; Worthy, Doug E. J.

doi:https://doi.org/10.5194/bg-13-1329-2016

Articles | Volume 13, issue 4

https://doi.org/10.5194/bg-13-1329-2016

© Author(s) 2016. This work is distributed under
the Creative Commons Attribution 3.0 License.

https://doi.org/10.5194/bg-13-1329-2016

© Author(s) 2016. This work is distributed under
the Creative Commons Attribution 3.0 License.

Articles | Volume 13, issue 4

Research article

|

02 Mar 2016

Research article |

| 02 Mar 2016

Evaluation of wetland methane emissions across North America using atmospheric data and inverse modeling

Scot M. Miller, Roisin Commane, Joe R. Melton, Arlyn E. Andrews, Joshua Benmergui, Edward J. Dlugokencky, Greet Janssens-Maenhout, Anna M. Michalak, Colm Sweeney, and Doug E. J. Worthy

Download

Final revised paper (published on 02 Mar 2016)
Supplement to the final revised paper
Preprint (discussion started on 23 Jun 2015)
Supplement to the preprint

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC C3815: 'reviewer comments on Miller et al. (BGC-12-9341-2015)', Anonymous Referee #1, 26 Jul 2015
- AC C6944: 'Reply to Reviewer #1', Scot Miller, 23 Oct 2015
RC C4229: 'Review of ‘The ability of atmospheric data to resolve discrepancies in wetland methane estimates over North America’ by S. M. Miller et al.', Anonymous Referee #2, 12 Aug 2015
- AC C6956: 'Reply to Reviewer #2', Scot Miller, 23 Oct 2015

Peer-review completion

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

ED: Reconsider after major revisions (05 Nov 2015) by Sönke Zaehle

AR by Scot Miller on behalf of the Authors (19 Nov 2015) Author's response Manuscript

ED: Referee Nomination & Report Request started (01 Dec 2015) by Sönke Zaehle

RR by Anonymous Referee #2 (21 Dec 2015)

Suggestions for revision or reasons for rejection

Review of ‘The ability of atmospheric data to resolve differences in wetland methane estimates over North America’ by S. M. Miller et al. November 2015

This revised manuscript presents an evaluation of a set of North American methane flux estimates from the global comparison project named WETCHIMP (Wetland and Wetland CH4 Inter-comparison of Models Project). This revision has addressed many questions and comments from the initial reviews. Currently the manuscript contains a more systematic analysis of all the WETCHIMP models using synthesis data Bayesian Information Criterion (BIC) experiments, atmospheric observations BIC experiments and WETCHIMP models comparison to atmospheric observations and inversion results. The additional explanations and clarifications have made the manuscript more understandable. However, some details are still missing and some questions remain (see specific comments below). Therefore this manuscript requires some minor revisions before it can be accepted for publication.

General comments

The results in the revised manuscript have changed substantially from the initial manuscript, specifically the results in Figs 3 and 4, as well as Table 1. This is an indication of the uncertainty in the experimental results, which depends strongly on several factors, including the number of WETCHIMP models used, the scaling or non-scaling of the EDGAR emissions, the scaling or non-scaling of HBL fluxes, and so on. Therefore the uncertainties and the lack of robustness in the analysis results need to be explained and discussed more clearly in the abstract, discussion and conclusion.

Specific comments

Page 2, lines 8-9: ‘we then use real data to analyze the magnitude, seasonality, and spatial distribution of each model estimate’, change ‘real data’ to ‘real data and inversion model results’ as some comparison used inversion results (seasonal cycle derived from real data through inversion analysis) rather than directly to ‘real data’. Inversion model results with potentially large uncertainties should not be confused with ‘real data’ or ‘atmospheric data’. The distinction between ‘real data’ and ‘inversion model results’ has to be clearly noted throughout the manuscript. In fact, the usage of inversion model results in addition to real data should be stated in the title of the manuscript.

Page 2, lines 10-11: ‘Many models predict a seasonality that is narrower than implied by atmospheric CH4 data’, change to ‘Many models predict a seasonality that is narrower than implied by inversion model results’

Page 3, lines 16-17: ‘The present study compares the WETCHIMP CH4 flux estimates against atmospheric CH4 data from 2007–2008 through two sets of analyses.’ Change ‘against atmospheric CH4 data’ to ‘against atmospheric CH4 data and inversion model results’.

Page 4, lines 24-25: ‘Based on these synthetic experiments, we conduct a second set of analyses using real atmospheric data.’ Change to ‘Based on these synthetic experiments, we conduct a second set of analyses using real atmospheric data and inversion model results.’

Page 6, lines 11-12: ‘H(nxm)’, define m.

Page 7, lines 18-19: ‘As a result, the model selection framework cannot scale other variables in X to reproduce the atmospheric CH4 signal from wetlands’ what are the ‘other variables’ referring to?

Page 7, lines 22-24: ‘For consistency among the synthetic datasets, we scale the annual HBL CH4 budget in each WETCHIMP model to match the overall magnitude estimated by several top-down studies’. Are the fluxes outside the HBL region scaled the same way? If the fluxes elsewhere are scaled differently than the HBL region, then each WETCHIMP model has been altered and the experiments are not evaluating the ‘original’ WETCHIMP models. How do the results vary with and without scaling HBL as mentioned?

Page 8, lines 3-5: ‘the intercept for each month is represented by a vector of ones in the matrix X, and this intercept is always included within X.’ The word ‘always’ is confusing. Is the ‘intercept’ included in ‘section 2.3 Real data experiments’ only or in ‘section 2.2 Synthetic data experiment’ also (then it should be mentioned in 2.2)?

Page 10, lines 4-5: ‘We run several additional test scenarios to explore why the synthetic observations may not always be able to detect wetland CH4 fluxes.’ How many ‘repeats’ were done in these test scenarios? How do the BIC test results vary with the number of repeats?

Page 12, line 22: ‘4 Results and discussion: comparisons with atmospheric data’ change to ‘4 Results and discussion: comparisons with atmospheric data and inversion model results’

Page 13, lines 19-20: ‘LPJ-Bern and LPJ-WHyMe also use land cover maps and/or land surveys to estimate wetland’ This is suggested as a reason why LPJ-Bern was ‘selected’ by the BIC experiment. Explain why LPJ-WhyMe was not selected and what other factors could be active.

Page 13, line 14: ‘all six WETCHIMP models’ change six to seven.

Hide

RR by Anonymous Referee #3 (04 Jan 2016)

Suggestions for revision or reasons for rejection

I was asked to review this study after the first round of revisions, which in my opinion improved the manuscript substantially. With a few additional (minor) modifications, as explained below, it should be ready for publication in BG.

BIC penalty term: Limited information is provided about the use of this term. To avoid confusion about it, some further sentences are needed explaining its role in the experiments that are presented. If I’m right it actually doesn’t play a role at all. The number of measurements is the same for every inversion and so is the number of explanatory variables (p=16 in all experiments if I am correct, it certainly doesn’t hurt to make this explicit in the text). Because of this the BIC penalty term is the same in every experiment. If this is correct, then please state this explicitly.

A definition is missing of the word ‘selection experiment’. It could either mean the entire procedure of selecting the optimal model or one single inversion. Page 7, line 6 suggests it is the latter (1000 times). Page 8, line 21 suggests it is the first. This should be clarified.

An explanation is needed of why the state vector is different for the synthetic and real-data model selection experiments. Strictly speaking, by introducing the monthly offset term we do not know anymore if results that were robust in the synthetic experiment are still robust in the real-data experiment.

Page 11, line 6: “The atmospheric network can better detect …”
This statement suggests that flux detection using atmospheric data benefits from good priors. I would rather argue that the poorer our priors, the more we can learn from atmospheric data. If the wetland models were an order of magnitude further apart, wouldn’t it have been easier to choose the right one?

Page 14, line 24: “Hence, the disagreement … global wetland budget”
What is suggested here is that overestimated emissions over the North America, also mean that global emissions are overestimated. However, this cannot be concluded from that fact that models that show high emissions over North America have large global emissions, in combination with evidence the North American emissions are overestimated. For that, additional evidence is needed that the global emissions are indeed overestimated. Otherwise global emissions could still be correct despite the fact that North American estimates were overestimated.

Figure S4: I’m surprised to see that the model is doing a rather poor job in reproducing the measurements at sites that are not much influenced by wetland emissions. To authors point to papers suggesting that EDGAR underestimates anthropogenic emissions over the USA. However, these differences point to an underestimate by about a factor 2, which seems too large. How about sites along the coastline, which directly sample the marine background? Do those measurements confirm that the initial concentrations are correctly dealt with? Some discussion is needed of whether the uncertainty of the anthropogenic emissions is well enough represented in the synthetic experiments. If the real data suggest that uncertainties are larger, then the derived detectability of wetland emissions may have been too optimistic.

A note of caution that is missing in the discussion section is that the WRF-STILT approach relies on an accurate representation of PBL mixing. Care has been taken only too select afternoon values. Nevertheless, it would be useful to have vertical profiles (e.g. from the aircraft measurements) supporting the point that is made in Figure 4.

Suppl, page 1, line 36: ‘At these’ io ‘At the these’

Hide

ED: Publish subject to minor revisions (Editor review) (12 Jan 2016) by Sönke Zaehle

AR by Scot Miller on behalf of the Authors (29 Jan 2016) Author's response Manuscript

ED: Publish subject to minor revisions (Editor review) (07 Feb 2016) by Sönke Zaehle

AR by Scot Miller on behalf of the Authors (12 Feb 2016) Author's response Manuscript

ED: Publish as is (17 Feb 2016) by Sönke Zaehle

AR by Scot Miller on behalf of the Authors (21 Feb 2016) Manuscript

Download

Article (1696 KB)
Full-text XML

Short summary

We use atmospheric data from the US and Canada to examine seven wetland methane flux estimates. Relative to existing estimates, we find a methane source that is smaller in magnitude with a broader seasonal cycle. Furthermore, we estimate the largest fluxes over the Hudson Bay Lowlands, a spatial distribution that differs from commonly used remote sensing estimates of wetland location.