Underestimation of boreal soil carbon stocks by mathematical soil carbon models linked to soil nutrient status

Ťupek, Boris; Ortiz, Carina A.; Hashimoto, Shoji; Stendahl, Johan; Dahlgren, Jonas; Karltun, Erik; Lehtonen, Aleksi

doi:https://doi.org/10.5194/bg-13-4439-2016

Articles | Volume 13, issue 15

https://doi.org/10.5194/bg-13-4439-2016

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

https://doi.org/10.5194/bg-13-4439-2016

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

Articles | Volume 13, issue 15

Research article

|

10 Aug 2016

Research article |

| 10 Aug 2016

Underestimation of boreal soil carbon stocks by mathematical soil carbon models linked to soil nutrient status

Boris Ťupek, Carina A. Ortiz, Shoji Hashimoto, Johan Stendahl, Jonas Dahlgren, Erik Karltun, and Aleksi Lehtonen

Download

Final revised paper (published on 10 Aug 2016)
Supplement to the final revised paper
Preprint (discussion started on 18 Jan 2016)
Supplement to the preprint

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

SC1: 'referee comments', Göran Ågren, 04 Feb 2016
- AC1: 'Author’s reply to Prof G. I. Ågren', Boris Ťupek, 16 Feb 2016
RC1: 'review', Leonid L. Golubyatnikov, 09 Feb 2016
- AC2: 'Author’s reply to Dr Leonid L. Golubyatnikov', Boris Ťupek, 25 Feb 2016
RC2: 'Review', Anonymous Referee #2, 07 Mar 2016
- AC3: 'Author's reply to Referee #2', Boris Ťupek, 09 Apr 2016

Peer-review completion

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

ED: Reconsider after major revisions (09 Apr 2016) by Alexey V. Eliseev

AR by Boris Tupek on behalf of the Authors (11 Apr 2016) Author's response Manuscript

ED: Referee Nomination & Report Request started (12 Apr 2016) by Alexey V. Eliseev

RR by Anonymous Referee #2 (22 Apr 2016)

Suggestions for revision or reasons for rejection

Thank you for your careful considerations of my comments from the first round however I have some of the same concerns from the first round and a few further points.

1. When I expressed confusion about the use of the term 'actual state' that should be a hint that maybe it is not a good term. It may be very common in physics, but this is not a physics journal, nor is the audience going to be primarily physicists. Use the language of your audience. Consider changing actual state to be 'observed' if the quantity is indeed observed. It is more clear then the derivation of the information. Just because it is the state at the present time doesn't tell me if it is an observed or modelled quantity. Similarly, I would use equilibrium value rather than steady-state. Again the language of your readers...

2. Why did none of the work with 'tuned' CENTURY not make it into the revised MS and only in the response to reviewers? That should go in the MS.

3. I don't agree that keeping soil group 8 with only 0.24% of points is valid. Either show that it is not just an artifact or error from your regression tree or remove it. It is obviously very strange given how much it differs from the other groups and how far away the models are. If it is a real group and not just sampling error or assignment error it should be better justified. There are 5 decisions steps to get to that group, prove that you didn't make a mistake on one of those five. Conversely really try to answer why that group is so special and look into the measurements to understand the conditions of those few sites better.

4. Fig B1, add the number of points on each (n=...) so us readers have that info.

5. I see now that Fig A1 caption no longer calls it 'measured fAPAR' but 'actual fAPAR'. So that seems to mean that we have 'actual fAPAR' which is a calculated quantity based on Harkonen et al. and your modelled fAPAR (?). So really we are comparing model vs. model? So where do things like observations fit in?

This brings me to a more major point. There is a heavy reliance on models to check models within this paper. For eg.

"We modelled the steady state biomass by applying the fitted exponential functions between the actual state forest biomass components (stem, branch, foliage, stump, coarse-roots, fine-roots, estimated by tree stand measurements and the allometric biomass functions) and the actual fraction of absorbed radiation (fAP AR ) (Appendix B) to the estimated fAP AR70 of the steady state forest" (line 146 in revised MS)

and then:

"Forest stand biomass was estimated by allometric biomass functions for stem with bark, branch,foliage, stump, coarse-roots and fine-roots applied to basic tree dimensions (breast height diameter,total height of tree, number of trees) of SFI stands " (line 130)

Then in appendix B:

"The biomass components derived with allometric models (measured) and those derived with fAP AR models (modeled) showed strong correlations (Fig. B1). "

But really they aren't measured! They are both modelled quantities; allometric models are not 'real'.

This seems like a rabbit hole where modelled quantities are used to check other modelled quantities which then generate further modelled quantities. How do you propagate error from the allometric relations? How do you know that using one modelled quantity to check another results in a correct answer? Since those are very hard questions to answer, it needs to be made completely transparent what is observed and what is modelled. So if you are comparing a modelled value to a modelled value there can be no mistaking it for measured quantities. What you have done could be valid but it is still very opaque on what is going on.

Hide

RR by Leonid L. Golubyatnikov (29 Apr 2016)

ED: Reject (04 May 2016) by Alexey V. Eliseev

ED: Reconsider after major revisions (18 May 2016) by Alexey V. Eliseev

AR by Boris Tupek on behalf of the Authors (18 May 2016) Author's response Manuscript

ED: Referee Nomination & Report Request started (20 May 2016) by Alexey V. Eliseev

RR by Anonymous Referee #3 (21 Jun 2016)

Suggestions for revision or reasons for rejection

General comments
The paper Ťupek et al. is aimed to test the ability of some biogeochemical models (Yasso07, Q, and CENTURY) to predict the soil organic carbon stocks in Swedish forests. For model validations a large set of forest and soil inventory data collected in different forest regions in Sweden are used. As one of the key results of the study it was shown that the models are not able to predict adequately the soil carbon accumulation in forest sites with very high nutrient status. The question about model ability to predict the soil carbon stock under over-moist soil conditions (high ground water level) was also discussed.
The main goal and objectives of the study is corresponded to the main scope of Biogeoscience journal and the paper can be published in BGS after some revision.
Before publication of the manuscript the several points have to be additionally clarified and discussed.
It is not completely clear why the authors use in the study "an actual fraction of photosynthetically active absorbed radiation (fAPAR) as a relative indicator of a site’s capacity to produce biomass"? Authors write that "the fAPAR was calculated using basic tree measurements". And from fAPAR values the forest productivity is estimated. Application of such indirect approach to estimate the forest productivity within the frameworks of the study is not clear. Actually the amount of above ground photosynthesizing biomass that characterizes the forest productivity can be either estimated directly from the forest inventory or derived from remote sensing data (e.g. from NDVI). It is clear that the fAPAR approach is broadly used in remote sensing to quantify the GPP and NPP of land surface in regional and global scales. However GPP and NPP estimations on local scale using this approach are usually very uncertain. Dependences of NPP and GPP on fAPAR even for growing non-damaged forest stands are often non-linear. The fAPAR is actually depended on many different parameters including amount of above-ground green biomass, amount of non- photosynthesizing biomass, forest understorey, soil properties including composition, texture and moisture, etc., etc.
Other point. Authors use the tree height as an indicator to quantity the forest "site" productivity. The forest productivity is actually depended on forest species composition, influenced by the air temperature, precipitation, soil properties, water and nutrient supply during the previous years, etc. Nowadays it is very popular to use the tree height as some relative indicator of forest productivity. However it is relatively rough indicator. In many cases the tree of different species reaches the maximal heights some time before the reference age used by authors - 100 years. In this case the tree height is not the best characteristics for forest productivity. The productivity (in particular aboveground production) of the trees for the time from the stage of seedlings to stages of mature or old trees can be actually characterized by two main characteristics - tree height and tree steam diameter (BHD). So it is better actually to use combined indicator for forest productivity including the tree height and BHD.
In figure A1 I suggest to replace the productivity (title of X axe) by tree height. To avoid any misunderstanding.
Specific comments
Line 3. "... carbon exchange ...". Do you mean the total carbon or CO2 only?
Line 26. Do you mean the carbon dioxide?
Line 331 Annual temperature cannot be never "cold". Cold winter, but the temperature is low or high.
Significance level (p=value) in Fig.4, Tab.A1, B1, C1 has to be indicated.
Table A1, B1. Please to round the numbers to 1 or 2 digits after comma.

Hide

ED: Publish subject to minor revisions (Editor review) (22 Jun 2016) by Alexey V. Eliseev

AR by Boris Tupek on behalf of the Authors (06 Jul 2016) Author's response Manuscript

ED: Publish subject to technical corrections (07 Jul 2016) by Alexey V. Eliseev

AR by Boris Tupek on behalf of the Authors (12 Jul 2016) Author's response Manuscript

Download

Article (2522 KB)
Full-text XML

Short summary

We evaluated the soil carbon stock estimates of Yasso07, Q, and CENTURY soil carbon models, used in national greenhouse gas inventories in Europe, Japan, and USA, with soil carbon stock measurements from Swedish Forest Soil National Inventories. Measurements grouped according to the gradient of soil nutrient status revealed that the models underestimated for the Swedish boreal forest soils with higher site fertility. We discussed mechanisms of underestimation and further model developments.