Interpreting eddy covariance data from heterogeneous Siberian tundra: land-cover-specific  methane fluxes and spatial representativeness

Tuovinen, Juha-Pekka; Aurela, Mika; Hatakka, Juha; Räsänen, Aleksi; Virtanen, Tarmo; Mikola, Juha; Ivakhov, Viktor; Kondratyev, Vladimir; Laurila, Tuomas

doi:https://doi.org/10.5194/bg-16-255-2019

Articles | Volume 16, issue 2

https://doi.org/10.5194/bg-16-255-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/bg-16-255-2019

© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 16, issue 2

Research article

|

22 Jan 2019

Research article |

| 22 Jan 2019

Interpreting eddy covariance data from heterogeneous Siberian tundra: land-cover-specific methane fluxes and spatial representativeness

Juha-Pekka Tuovinen, Mika Aurela, Juha Hatakka, Aleksi Räsänen, Tarmo Virtanen, Juha Mikola, Viktor Ivakhov, Vladimir Kondratyev, and Tuomas Laurila

Download

Final revised paper (published on 22 Jan 2019)
Supplement to the final revised paper
Preprint (discussion started on 10 Apr 2018)
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: 'Reviewer comment to Tuovinen et al.', Anonymous Referee #1, 03 May 2018
- AC1: 'Response to Referee 1', Juha-Pekka Tuovinen, 20 Jul 2018
RC2: 'Review of Tuovinen et al', Anonymous Referee #2, 05 Jul 2018
- AC2: 'Response to Referee 2', Juha-Pekka Tuovinen, 20 Jul 2018

Peer-review completion

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

ED: Reconsider after major revisions (03 Aug 2018) by Lutz Merbold

AR by Juha-Pekka Tuovinen on behalf of the Authors (21 Sep 2018) Author's response Manuscript

ED: Referee Nomination & Report Request started (04 Oct 2018) by Lutz Merbold

RR by Anonymous Referee #2 (17 Oct 2018)

Suggestions for revision or reasons for rejection

This is my second review of the paper by Tuovinen et al. and although I feel that the manuscript has improved, I think it still needs a lot of work because there are many aspects that are simply not clear enough. What the paper aims to do is not overly complicated, but the text is still very difficult to follow from the methods section onwards. Moreover, the approach relies strongly on idealized mathematical theory with a long list of assumptions, which disregard the chaotic nature of natural ecosystems. These assumptions may be valid for the purpose of this paper but need to be better supported.

Section 2.2.1 reads as a background chapter in a textbook without a clearly presented goal. The frequent use of “we can” without a clearly stated purpose does not make it obvious to the reader how these equations will be used in the rest of the paper. A reader may wonder whether this is background information or whether it’s used in the final calculations. This is not clear until the reader has reached the results section. It would help to rewrite the methods section and clearly indicate at each stage what the goal is of the chosen approach, without ambiguous wording. This helps the reader to follow the paper a lot better and avoids that people will have to read it twice to understand the approach.

The same unclarity goes for the many assumptions this paper makes in the methods section, which are not properly argued or followed up on. Here is a short list:

- In section 2.2.3 sigma_v/u* is fixed at the median value of 2.3, but what is the consequence of doing this for all stabilities? By how much does sigma_v/u* vary and does it change your results?
- Page 10, line 16: it is assumed that methane fluxes do not vary within an LCC which is very unlikely considering previously published studies. The effect of this choice is not shown.
- Page 10, line 20: it is further assumed that LCC-specific fluxes remain constant in time even though Figure 6 shows this is not the case.
- Page 11, line 13: it is assumed that these confidence intervals reflect the overall uncertainty, but errors in ec-fluxes are highly non-linear. Is OLS still applicable in that case? Also, no evidence is presented that this is a proper replacement for a bottom-up error analysis.
- Page 11, line 21: it is assumed that model residuals represent measurement error. This doesn’t make sense. These residuals indicate how well your model performs, not the other way round. Yes, there is uncertainty in ec-measurements but there are well-established methods in the community on how to calculate random and systematic measurement errors.
- Page 12, line 11: it is hypothesized that the neutral group is not statistically different from zero. Is this tested? Figure 6 appears to suggest otherwise, especially in week 7.

All these assumptions explain why there is such a huge difference in the spread between the modelled and measured values. This does not represent, as suggested by the authors, measurement error, but rather an inability of their model to represent this variation. This should be easily visible from either a scatter plot or a time-series, but this is not presented.

The lack of a correlation with soil temperature, as mentioned on page 15, line 24, and page 7, line1, may have to do with the fact that soil temperature was measured at a single point, although this is not clearly stated in the paper. If so, it’s unsurprising that there is a poor correlation with spatially averaged data. Also, it’s not clear to me whether this was a linear correlation? Methane fluxes vary non-linearly with temperature. Furthermore, Figure S1 clearly shows that the model and soil temperature have highly different time intervals. Does the correlation become better when soil temperature is binned according to the same intervals as the model?

But in principal, the investigation of the temporal variation in fluxes should be done at a higher frequency. Methane fluxes can vary strongly within a day and within a week, especially after rain storms. However, the influence of changes in the water table on temporal trends is not discussed in this paper. The calculated TWI provides only spatial, but no temporal information.

Finally, I’d like to comment on my earlier review where I said that adjusting the study area to achieve a better model fit “felt like cheating”. I can understand how this may have been perceived as an accusation, which was not my intention. I could have used less candid words, but my remark was an honest response to the way this was written. If post-hoc changes are made to make the results look better, this does not look favorably. If I, as a reviewer who tries to read this paper as carefully as possible, misinterpret what’s written then others will do so, too. It’s important that the authors avoid such misinterpretation. In that light, the addition of the extra sentence about post-hoc interpretation is welcome. If the authors also add an extra reference to Figure 7 for clarification, it may further help to avoid misinterpretation by others.

Minor comments:

The authors insist of the use of microgram/m2/s as a flux unit, since gram is an SI unit. This was not the point of my remark. CO2 fluxes measured with ec are almost always presented as micromol/m2/s and within the flux community it has become common practice to report methane fluxes in nmol/m2/s. The use of grams is a remnant following from flux chamber studies but not helpful in the context of eddy covariance unless cumulative budgets are discussed.

Page 12, line 5: which other fens?

Page 12, line 6: which previous data?

Page 14, line 4-10: the studies cited here used very different scales for their analysis. How is this a comparison?

Page 14, line 22: is this map derived in the same way as the smaller map?

Page 15, line 9: the inability of your model to capture high-frequency dynamics in methane fluxes should not be dismissed as ‘measurement noise’

Page 20, line 2: the poor fit to the high frequency data, as shown in figure 5, shows that this method cannot be used for gapfilling of the time series at 30 min intervals. It only provides averages which may make sense at seasonal timescales, but not shorter. There are much better gapfilling methods available.

Hide

RR by Anonymous Referee #1 (25 Oct 2018)

Suggestions for revision or reasons for rejection

Review on the revised version of the manuscript titled ‘Interpreting eddy covariance data from heterogeneous Siberian tundra: land cover-specific methane fluxes and spatial representativeness’ by Tuovinen et al.

I will restrict this review to the revised aspects of the newly submitted manuscript, and focus on the changes that address my major points of concern from the previous review. Overall, I acknowledge that my major areas of concern have been addressed well by the authors, so that the current version of the text has been significantly improved. I still think there is room for improvement, but this may be covered in a follow-up study that was announced by the authors in the conclusions.

Structure and length of the manuscript text are now well chosen, here particularly the methods section has benefitted from the revisions compared to the original manuscript. The results and their interpretation, separated into 3 major blocks, are convincingly organised, and well connected. The presentation of results in figures and tables is adequate. As the bottom line, I have only a few minor comments that should be addressed before publication of this study (see below).

Besides these minor things, one of my major concerns, the assessment of uncertainties and their impact on the study results, still has not been fully addressed. I acknowledge that a new section (3.4) has been added discussing methodological issues, but this is far from being a comprehensive uncertainty assessment. Most of the things that are addressed here in a qualitative fashion could as well be quantified, e.g. by running repeated optimisations based on randomly disturbed LCC maps or the input parameters for the footprint model. Also, the uncertainty of the eddy fluxes and their impact on the performance of the flux partitioning should be treated. I believe that the presented manuscript could be substantially strengthened through such a quantitative uncertainty assessment; however, I am not demanding this as mandatory for the given paper. If the authors decide to leave their uncertainty treatment at a qualitative level for now, I highly recommend to include the quantitative assessment for the follow-up paper.

minor comments:
- the 4th paragraph of the introduction (p.3 ll.10ff, general intro on footprints) is not contributing much to the story. It may be shortened, and combined with the following paragraph
- move 2.3.2 further up, since 2.3.1 partly refers to it. Options would be to either either swap 2.3.1 and 2.3.2, or move this up to the landscape classification section as 2.1.5
- in Section 3.2, the only part missing is that the actual LCC fluxes are not given in the text (only in a table), so that a quantitative comparison against chamber data does not make sense
- p.20, l.9ff: it is unclear how the discussion on definition of target area adds to this story (same for the conclusions)

Given these minor edits will be implemented, I recommend the manuscript for publication.

Hide

ED: Publish subject to minor revisions (review by editor) (08 Nov 2018) by Lutz Merbold

AR by Juha-Pekka Tuovinen on behalf of the Authors (22 Nov 2018) Author's response Manuscript

ED: Publish as is (08 Dec 2018) by Lutz Merbold

Download

Article (3323 KB)
Full-text XML

Short summary

We analysed ecosystem-scale measurements of methane exchange between Arctic tundra and the atmosphere, taking into account the large variations in vegetation and soil properties. The measurements are spatial averages, but using meteorological and statistical modelling techniques we could estimate methane emissions for different land cover types and quantify how well the measurements correspond to the spatial variability. This provides a more accurate estimate of the regional methane emission.