the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Mar 2025
Carbon soil stock change in an intensive crop field near Paris reveals significant carbon losses
Abstract. Soil is a large pool of carbon (C) storing globally twice and three times more carbon than the atmosphere and vegetation, respectively. Soil organic carbon (SOC) stocks are significantly impacted by land use changes, either negatively when forest or grasslands are turned into crops or positively when the opposite is done. This idea has led to the 4per1000 initiative to increase carbon storage in soils. However, intensive cropping and climate change may lead to organic and inorganic carbon losses from soils, which calls for long-term observations of soil organic carbon stocks in reference ecosystems over the globe. This idea is behind developing a reference soil sampling protocol for all EU Integrated Carbon Observing System research infrastructure (ICOS) ecosystem sites. We present the first case study of SOC stock measurements with the ICOS protocol at the French crop site FR-Gri in 2019 and compute the soil stock evolution from 2005 to 2019 for a wheat-maize-barley-oilseed-rape crop rotation. A significant decompaction of the 0–15 cm soil layer was observed over the 15 years, with a 25 % decrease in bulk density in the 0–5 cm layer and a 10 % decrease in the 5–15 cm layer, which can be attributed to reduced tillage performed since 2004. A significant increase of organic carbon content was observed in the same soil layers, around 15 % in the 5–15 cm layer and 20 % in the 0–5 cm layer. Despite its higher SOC content, the carbon stock decreased significantly in the 0–5, 5–30 and 0–60 cm layers because of the decreased bulk density. Overall, the SOC stock decreased by around 960 g C m-2 over the 13.25 years in the 0–60 cm layer, which corresponds to a decrease rate of 72 ± 17 g C m-2 yr-1 as estimated with a fixed depth approach. The equivalent soil mass approach led to a very similar estimate of 70 ± 16 g C m-2 yr-1. Overall, an organic carbon loss of 0.6 % year-1 was observed, consistent with previous studies. We utilised the soil carbon cycling model AMG to simulate the soil carbon dynamics over the period and beyond. Based on recorded exports and imports and estimated residue return to soil, the model was in perfect agreement with the soil stock measurements. The AMG model simulation was also consistent with the carbon flux balance reported between 2006 and 2010 as reported by Loubet et al. (2011), and predicted a stabilisation of the soil stock around 2028 with an overall decrease of 15 % of the soil organic carbon stock compared to 2005. This observed destocking may be explained by a shift towards larger exportations and lower residue return at this site compared to previous practices. This result questions the feasibility of the 4per1000 aspirational target and stresses for high-quality soil SOC evolution measurements as developed by the ICOS research infrastructure to answer this question.
- Preprint
(835 KB) - Metadata XML
-
Supplement
(148 KB) - BibTeX
- EndNote
Status: open (until 04 May 2025)
-
AC1: 'High quality figures to ease the reader', Benjamin Loubet, 21 Mar 2025
reply
We realise that the quality of the figures in the preprint was difficult to read.
To make this study easier to read, we are therefore proposing a supplementary document containing the figures in a better quality format.
-
RC1: 'Comment on egusphere-2025-592', Moritz Laub, 23 Apr 2025
reply
In this study, Loubet et al., describe two measurement campaigns in the FR-Gri site, which studies SOC stocks in 2005 and 2019. The also report differences in SOC between both time points, and simulate the evolution of SOC in 0-30 cm, using the AMG model.
Overall, the study is an interesting description of how SOC stock changes could be assessed by repeated samplings and the results are worth publishing. The overall context is well described, the authors describe clearly what they did (except the AMG model) and I think the results hold value. However, in the present form there are several major issues that the authors neglect. The first is the differences in sampling design between 2005 and 2019. This should be addressed explicitly. Using only a subset of 2019 samples might be the way to deal with it, but it must be ensured that the subset is representative of the 2005 locations, which is not demonstrated in the current article. The uncertainty arising should be addressed in the discussion. Additionally, the authors should try their best to harmonize the 2005 and 2019 sampling campaigns. This means that the horizons should be adjusted to represent the same depths/soil masses, for example by weighted means and by using an ESM quadratic spline for the stocks (Wendt and Hauser, 2013). Finally, the discussion should try integrating the results with other studies on a deeper level than what is currently done. Just stating they found X, we found Y, without much discussion why this could be, is a bit shallow.
I think if these issues are addressed properly, the study will make an interesting article.
Major concerns:
- The image quality of all figures is very bad. A higher resolution is needed to make them acceptable and readable!
- While the study acknowledges that the equivalent soil mass (ESM) approach is the much better one, compared to fixed depth, this does not show in the results. The fixed depth results are always reported first and discussed in more detail. If anything, the ESM results should be reported first, and the fixed depth approach only briefly mentioned, knowing that BD changed and it is therefore biased.
- The areas are so different that it is hard to state whether the difference between SOC in 2005 and 2019 is due to sampling design or real change in SOC!
- The same is for the different depths and soil layers sampled. In a way this article compares apples to pears without trying to make adjustments (e.g., by harmonizing the data to have the same depth intervals).
- The uncertainties that arise from this should at least be highlighted in the results (proper wording, lower and higher instead of decrease and increase) and the discussion (state the uncertainties explicitly).
- It is not clear how the “reduced” sampling area was chosen. In Figure 4, it says it was chosen to have the same fine earth fraction as the 2005 sampling. More details are needed in the methods, as this is a potentially influential choice for the rest of the study. And why was it not choosen based on spatial proximity?
- Any results of SOC stocks with the fixed depth approach should be reported with a high level of caution! By stating that the BD was significantly different between the samplings, you practically invalidate them and I would almost suggest that you completely eliminate them. If you want to report SOC stocks for different layers, then use the approach of Wendt and Hauser, fitting a quadratic function of cumulative SOC stocks to cumulative soil mass (a+ bx+cx^2) with 0 intercept per each auger point, make sure that the R² is close to 0.99 for each, and then predict the SOC stocks for the layers that you want. Then do statistics on these, which ensures that you are now comparing data that is normalized to the same equivalent soil mass and you “compare apples to apples”.
- The use of the AMG model is an interesting addition to the study. However, it is not well integrated with the rest of the text. It should be clarified in the introduction why it was used (I guess to compare input/output based SOC stock changes to measurements). Also, much more detail, such as parameters, inputs, etc. are needed in the methods. And why not use the AMG model to explore other management options that export less of the straw?
- The discussion only loosely connects the results of this study to those of others (we found X, they found Y). A more integrative discussion is needed, discussing what the results of this study mean in the context of the others, and what the novelty of this study is.
Specific comments:
L25+27 Would be good to state initial values of bulk density and of SOC
L31 Since you indicate a change in in bulk density, I think you should only present the results based on equivalent soil mass! The other ones are not right, that is why you used the equivalent soil mass approach.
L32 It must be 0.6% of the initial carbon not % of soil carbon. Please clarify.
L35 Should you not evaluate the change in SOC stock since 2005 rather than absolute SOC stock?
L40 While I agree that 4pm is probably unrealistic, you should not claim that the results from a single field are a good disproof.
L50 Should you not cite the Sanderman paper here as a more recent estimate of soil C loss?
L58 Evaluated? Where is the proof? Or do you mean postulated?
L69 Large uncertainty is an understatement if the confidence interval includes the 0 within less than one standard deviation.
L82 And, even worse, if the PTF includes SOC, as it often does, SOC and BD measurements are not independent. This can then lead to systematic errors.
L107-109 Since you will only evaluate one site, I think the two sentences about sampling design are a bit off-topic. To be removed, potentially.
L123 It is the long-term monitoring site in Gringnon, right? I would at least spell out the name once, as it is widely known.
Figure 2 Why is strata 9 excluded in 2019? And why are strata 7 and 10 included? They are much further from the 2005 samplings. If anything, you should eliminate 7 and 10 and include 9.
L210 Would be better to show all factors that you multiply with to do unit conversion than a single value that you collapse them to.
L250 A slightly better approach would be to follow the approach of Wendt and Hauser (2013), fitting a cubic regression of cumulative SOC stocks vs cumulative soil mass. As I understand it, your approach assumes a linear change of SOC stocks with soil mass, which may not be the best representation. Given that most soil stocks are in the topsoil, I guess it is not a huge error that you get by this, though.
L268 Did you include all strata? Or just the ones that overlap with the 2005 sampling campaign? The latter would be preferable since the 2005 area covered is much smaller.
L300ff A lot of details missing on the AMG model. How was management simulated? Were plants simulated? What allometric equations were used to estimate root inputs (i.e., what percentage of NPP went into roots)? Where and how can the AMG model be assessed? Which parameterization was used?
L315 The sample name SP-15 is not meaningful for external readers. I suggest you define it.
L317-319 With such a different sampling design I would be careful to infer temporal changes such as “decrease of bulk density” and just state that it was lower in the 2019 campaign.
Figure 4 It is not a meaningful displaying of the data if you mix all soil horizons into one graph, especially, given that they had different layers and sampling depths (60 vs 100 cm). I think to make any meaningful comparison, you HAVE TO display per depth horizon. E.g., the 0-5 cm horizons is the same, the 30 to 60 horizons in 2005 samples can be combined by computing the mean, and 5-15/15-30 by a weighted average.
Also, the strategy to choose the “reduced” sampling area just on having the same fine earth fraction as the 2005 sampling seems suboptimal. Why not choose based on spatial proximity?
L338 Again, “decreased” should be “was lower” since the sampling positions were not really the same. This wrong wording exists through the rest of the text, but I will not mention it each time. Please try to correct throughout.
L 339ff Please add p values to your statements of significance
Figure 5 I think it would be best to display this data as dots with error bars for your confidence intervals. The mean depth of each horizon could be used to display the dot and dots should be connected. At the moment it is really hard to read. Also, can you use annotations to show where there were significant differences?
Also figure 5 Why are the SOC stocks of 2005 missing in 30-60?
L348 For the SOC content you use the word “higher” to compare, which is the right way to address it.
L352 What is slightly significant? p < 0.1? Please add p values.
L353 What “correlation” do you mean? If you specify correlation, please give the r, also say if it is Pearson or Spearman. From the Figure S2, it does in fact not look like a strong correlation at all.
L357-364 Please do not use words such as “diminished” which imply temporal connection. Especially, when you talk about equal depth intervals. Focus on your ESM results, which is the only correct way to view it, given the BD has changed.
L375 This sentence is discussion. Also, what makes you so sure that the difference is not due to the different sampling design?
Figure 6 I suggest showing both 0-30 and 30-60 in this figure but ONLY with the ESM approach. You may display the fixed depth approach in the supplement.
Figure 7 I suggest you replace this figure by ESM based SOC stocks, based on soil masses (that you report together with the cm) that represent roughly the depths you want to show. Simplest way to do this would be Wendt and Hauser, that you also cite.
L431 I would start the discussion with a statement based on your data, and only then follow with the other studies in the following sentences.
L436 Again, if you apply the Wend and Hauser approach, you could easily estimate SOC stock differences for an ESM representing 0-15cm.
L441 Not necessarily. How much more robust 0-60 is really depends on how much BD changes and on how strong your SOC gradient is with depth (as you correctly argue in L455). ESM is the standard!
L450 As long as you do not use equivalent soil masses, your comparisons of 0-5cm etc. layers are really kind of meaningless!
Chapter 4.1 could be significantly shortened. A lot of the discussion is around what we already know – ESM is the better approach. Using the ESM approach for all depth/soil mass layers would really help to improve the discussion. E.g., are the differences in the upper horizons still there, if ESM is used.
Chapter 4.2 Can you really conclude that you have a change in SOC stock with the difference in sampling design? Is the MDD of Schrumpf et al. (2011) already considering a different sampling design, or is it meant only for the exact same sampling design?
Chapter 4.3 It would be important to state for any study that you report, whether is was ESM or fixed depth based. Overall, too long as well. The results of all the other studies could be summarized much more concisely.
L487 Why the switch to % losses from absolute values?
Chapter 4.4 is the best discussion chapter. It really tries to dig deeper and explain. The other chapters also provide room for that. For example, it could be discussed in more detail why there is a need to use ESM based on your and other data, and what the error is using a fixed depth approach.
L520 What do you mean by “it is excluded that the change comes from a change in crop cultivation”. Please clarify.
L532 Please mention where was the Leifeld study was conducted. Switzerland?
L553 N per ha and YEAR?
Chapter 4.5 While interesting, the part on inorganic carbon leaching is mostly speculation. Your study adds little here so this part should be kept to a minimal. In contrast, you MUST include a paragraph about the uncertainty that arises in your study due to the 2005 and 2019 sampling campaigns using different sampling designs! This uncertainty may reduce if you use ESM and discuss the error reduction properly.
Conclusions The uncertainties of the approach should be mentioned.
L578 Rather demonstrating the need to use ESM! (But increasing sampling depth is also good)
Citation: https://doi.org/10.5194/egusphere-2025-592-RC1
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 173 | 26 | 8 | 207 | 10 | 6 | 8 |
- HTML: 173
- PDF: 26
- XML: 8
- Total: 207
- Supplement: 10
- BibTeX: 6
- EndNote: 8
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| France | 1 | 73 | 33 |
| United States of America | 2 | 51 | 23 |
| Germany | 3 | 19 | 8 |
| China | 4 | 14 | 6 |
| Belgium | 5 | 7 | 3 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 73