the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 30 Apr 2026
Can secular stable soil organic carbon be isolated? An assessment of Zimmermann fractionation using a long-term bare fallow
Abstract. Useful soil organic matter fractionation techniques require fractions with accurate and distinct residence times and must be implementable within reasonable timeframes and with reasonable equipment. Among these techniques, the Zimmermann fractionation is a size- and density-fractionation process which aims at separating soil organic matter into five fractions known to be linked to kinetic compartments of carbon, including one fraction corresponding to the organic carbon in silt and clay fractions resistant to chemical oxidation (rSOC) considered to be stable at the scale of millenia, which is challenging to verify. In this paper, we used a 97-year-old long-term bare fallow experiment located at Versailles (France), to assess the stability of this fraction at the centennial scale. We applied the Zimmermann fractionation to archived soil samples collected in 1929, 1962 and 2021. We observed that the rSOC content was nearly constant over the lifetime of the trial, in contrast to total SOC and to other fractions. However, the rSOC only accounted for a small proportion of total SOC and did not contain all soil carbon persisting at the centennial time scale in this experiment. Our results show that rSOC represents an almost pure but not exhaustive pool of centennially stable SOC and that other methods are needed to obtain a complete estimate of centennially stable SOC.
- Preprint
(824 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-1028', Anonymous Referee #1, 09 Jun 2026
-
AC1: 'Reply on RC1', Amicie Delahaie, 07 Sep 2026
We warmly thank the referee for their constructive comments. Please find below our responses in italics.
This study aims to explore the stability of various SOC fractionation using a centennial-scale resampling approach on a bare fallow site to. Research that tests methodological reliability like this is instructive for both our understanding of carbon cycle processes and the optimization of carbon models. However, the core conclusions remain open to question:
The authors infer the persistence of rSOC based on changes in carbon content over time. In a bare fallow experiment with no carbon inputs, this approach may be partially valid. However, the authors acknowledge an unquantified black carbon input during World War II. Black carbon is not absolutely stable. Microorganisms can process it into short-chain compounds and microbial products, which may subsequently contribute to rSOC and other fractions. Additionally, other SOC fractions can also be a source of rSOC. Therefore, inferring centennial-scale persistence solely from content changes is insufficient. The key conclusion requires more direct evidence, such as radiocarbon (Δ14C).
Radiocarbon analyses are currently being conducted in an collaborating laboratory (CEA-LSCE, Gif-sur-Yvette, France). We planned to conduct them but not to include them in the paper; however, following your recommendation and those of Referee #2, the results will be added to the manuscript in order to strengthen - or invalidate - our conclusions.
The reported mass recovery rates are implausibly high, reaching over 200%. Such anomalies compromise the reliability of results.
Please see detailed answer below.
Some results are based on subjective visual inspection (Figure 4), which is questionable.
Please see detailed answer below.
Together, the current evidence is not sufficient to support their core conclusions. The detailed comments are as follows:
Title：I suggest explicitly stating “centennial-scale” rather than “long-term.”
We will modify the title to clarify the timescale.
L27：I recommend that the authors insert a paragraph break, so that the meaning of the fractionation is introduced first, followed by a description of the specific techniques used.
The text will be modified accordingly.
L27-38：The section opens with the assertion that physical fractionation is more reliable than chemical methods. However, the majority of the paragraph is devoted to thermal analysis techniques. This structural imbalance distracts from the core message and may confuse readers about the primary focus of the section.
We will modify this paragraph to better focus on the comparison of physical vs chemical methods. However, we find it important to still present thermal analysis, as these methods have been investigated for decades to virtually fractionate OM (see Balesdent 1996, 10.1111/j.1365-2389.1996.tb01848.x); we will better justify their presentation as well.
L39-45：I am curious whether the five Zimmermann’s fractions correspond one-to-one with the RothC pools. The statement in L45-46 seems to imply such a strict correspondence. However, the BIO pool in RothC is difficult to characterize using the Zimmermann fractions. It may be more closely related to MBC instead.
The Zimmermann fractions do not - and are not meant to - correspond one-to-one with the RothC pools. Instead, Zimmermann et al. (2007; 10.1111/j.1365-2389.2006.00855.x) use calculations at the equilibrium in RothC to establish a correspondence between RothC pools and the physical fractions through splitting ratios, except for the rSOC fraction which is indeed meant to correspond directly to the IOM pool.
We attached in Supplement a figure from Poeplau et al. (2013; DOI: 10.1111/ejss.12088) showing the relationships between the pools and fractions (the fraction ‘s+c-rSOC’ from the figure corresponds to the sSOC fraction in our paper).We will clarify this part to avoid suggesting any direct equivalence between those.
L74-82：How were the soil samples preserved? For a centennial-scale experiment, preservation conditions are critical to the reliability of the measurements.
The samples were sieved after sampling and air-dried for at least one week, then kept in glass jars in a dark storage room with no regulation of the temperature. Numerous analyses were already conducted through time; specifically, the samples of specific years (e.g. 1977 samples) were re-analysed multiple times several years apart (e.g. in 1977, 1991, 2002, 2014, 2021), and the results show no alteration of the organic matter during storage.
L104：Does ultrasonic dispersion carry a risk of aggregate disruption, potentially leading to an underestimation of the S+C fraction?
I think you mean the S+A fraction. If aggregate disruption indeed occurs, S+A will be underestimated while s+c will be overestimated.
This is indeed an important point, and it has been stressed out by Poeplau et al. ((2013; DOI: 10.1111/ejss.12088). In their paper, they had the Zimmermann fractionation process completed by several labs in order to highlight interlaboratory differences in the results and identify critical worksteps that need to be carefully harmonized in order to obtain meaningful, reproducible results. In particular, they highlighted a strong influence of ultrasonic dispersion on the results. The problem they pointed out is that while the (volumic) energy is prescribed in the Zimmermann protocol (22 J.ml-1), what matters is also the power. The six laboratories that completed the Zimmermann fractionation used power ranging from 22 W to 90 W - the time was set accordingly to match an energy of 22 J.ml-1. The results of this ring trial showed that the lab using the smallest power (22 W) had the best results, so Poeplau et al. recommended a power of 20 W. The power of our ultrasonic probe was checked by linear regression before every use and was between 21 and 23 W, very close to the recommended power. We are thus confident that our S+A and s+c fractions are correctly estimated, according to Zimmermann et al. protocol and Poeplau et al. recommendations.
L138-142：The elemental analyzer measures the carbon content per unit mass of each fraction, but the mass of each fraction differs. Has the carbon content been normalized to that per unit mass of air-dried soil?
Yes, the carbon content has been normalized. In the text, values are given either in mg C g-1 of soil or in mg C g-1 of fraction, depending on what we want to highlight.
L146：Major concern! The mass recovery rates in 1962 and 2021 are anomalous (in some cases exceeding 200%). This seriously undermines my confidence in the reliability of the results. This means that the authors started with 30 g of air-dried soil but ended up with 60 g of recovered soil fractions? This is physically impossible and difficult to understand.
Based on the mean mass recovery (98.16% to 100.05%, L145), I suspect that only a few samples have anomalous recoveries. I respect the authors’ transparency in reporting all data, but I recommend that they conduct a sensitivity analysis by excluding the samples with excessively high mass recoveries to test the robustness of results. In addition, reporting the carbon recovery rates is also necessary.
The mass yields do range from 98.16 to 100.05 % (L145). These are not mean values: all mass yields, for all samples and all years, do fall within these values.
The anomaly lies only in the carbon recovery rates, which indeed reach abnormally high values. The carbon balances are already displayed L145-146: “The carbon yield ranged from 92.45 to 100.10 % for 1929 samples, from 109.94 to 231.22 % for 1962 samples, and from 80.53 to 176.81 % for 2021 samples”.
The abnormality in carbon yields is indeed concerning, but the fact that all our mass yields are very close to 100% reassures us that we manipulated correctly. The explanation we could come up with is the unexpected black carbon found in 1962 and 2021: this black carbon, which would be contained in the POM fraction, has a very high C content, which means that even the slightest error in mass - which would not affect the mass balance - would have a great impact on the carbon balance.
L175-180：The authors rely solely on visual observation to infer the source of POM and temporal changes in black carbon content. Such subjective analysis is inappropriate for presentation as formal results. Strictly, carbon source should be based on quantitative methods such as biomarker or molecular fingerprinting (e.g., 13C-NMR, Py-GC/MS). While I understand that the authors may lack the resources for such analyses, presenting subjective visual observations as evidence is scientifically unrigorous.
The request for more in-depth analysis of this black carbon is fully understandable. The radiocarbon results will be incorporated in the manuscript as soon as possible, as requested both by yourself and Referee #2. We can also consider adding 13C-NMR, Py-GC/MS, CTO-375 or other complementary analyses to identify and measure black carbon.
However, the visual observations were not our only clues to infer the nature of POM. Section 3.3 details sudden changes in the C content of the POM fraction, with high contents (717.4 ± 28.4 mg C g-1 of fraction) consistent with the ones found in black carbon. Also, the C:N ratio doubles between 1929 and 1962, reaching 64.2 ± 4.1, which is another feature found in black carbon. These features pointed to the presence of black carbon, and this hypothesis was consistent with visual observation.
Discussion: The discussion would benefit from explicitly linking the findings back to the RothC model. If rSOC represents only a portion of stable SOC, then using rSOC as a direct proxy for IOM pool in RothC may systematically underestimate the soil carbon sequestration potential.
We understand this concern regarding the RothC model in view of our results. However, our work is only based on a 98-years-old bare fallow, which means our experimental results can only cover such a timespan, rather than 10,000 or 50,000 years as described in the model. For this reason, we chose to focus only on the potential of rSOC to represent a centennially stable fraction rather than on verifying Zimmermann fractionation’s ability to correctly initialize RothC compartments.
However, our results indeed suggest that IOM and rSOC are highly unlikely to correspond one-to-one. We will make it appear more clearly.
Figure 2：Please clarify the definition of sensitive and resistant SOC.
The sensitive SOC (subject to intense oxidation) refers to sSOC, and the resistant SOC (subject to the same oxidation) refers to rSOC, as defined in the manuscript, that is, the two complementary parts of the s+c fraction. Operationally, the resistant SOC is what is left from s+c after oxidation, the sensitive SOC is the part that is oxidized.
We will state it more clearly in the caption.
-
AC1: 'Reply on RC1', Amicie Delahaie, 07 Sep 2026
-
RC2: 'Comment on egusphere-2026-1028', Anonymous Referee #2, 13 Jul 2026
In their manuscript, titled ‘Can secular stable soil organic carbon be isolated? An assessment of Zimmermann fractionation using a long-term bare fallow’, Delahaie et al. present their results of the fractionation of 15 soil samples (collected at three points in time (1929, 1962 and 2021) in five replicates) at the “42 plots” experiment in Versailles (France). The samples were collected from control plots of the long-term bare fallow experiment, which did not undergo any treatment. The samples were fractionated using an adapted version of the Zimmermann fractionation to test the validity of the assumption that the residual SOC fraction (rSOC) remains stable on a centennial timescale. The latter is a necessary condition for the assumption of the RothC model, which assumed its inert organic matter pool has a residence time of 50,000 years.
The authors found that the OC content of the rSOC pool slightly decreased, from 0.66 mg C/g soil in 1929, to 0.55 mg C/g soil in 2021. Based on this, the authors conclude that this fraction predominantly consists of OC with an at least centennial residence time. In addition, as there is a considerable amount of SOC present in other pools in the samples collected in 2021, the authors conclude that not all centennially-cycling SOC is present in the rSOC pool. Besides these main results, the authors discuss the evolution of other isolated fractions, for example why the C concentration of the POC is highest in 1962, and how the presence of coal affected the measurements.
The published articles I mention in my feedback have been chosen based on their scientific relevance, and I leave it up to the authors whether they want to include these in their manuscript or not.
General feedback
The topic of the manuscript, quantifying similarities between measured and modelled SOC fractions, is very relevant. The manuscript is well-written, logically structured and the figures clearly visualize the results.
The main message, that the rSOC fraction from the Zimmermann fractionation represents highly persistent SOC at the centennial time scale, is based on the fractionation of 15 soil samples collected at three points in time, on which only the C and N concentration was measured. While the approach of fractionating archived soil samples spanning ca. 90 years is highly welcomed, this is very small number of samples and analyses to draw the conclusion upon. For example, it is not clear to me why the following approaches were not considered or applied:
- Measurements of 13C or 14C isotopes. Based on the results, the authors can only conclude that the C concentration of the rSOC fraction didn’t change, but not that its composition remained stable (i.e., that the rSOC that was present in 1929 is the same as measured in 2021, while no C was transferred to rSOC from other pools). While the relatively stable C:N ratio of rSOC through time hints at a stable composition, the authors could make a much stronger case by analysing more characteristics of the rSOC fractions and assessing if these remained stable through time.
- Fractionation of samples from other treatments. The result show that over a timescale of ca. 90 years, also fractions that are supposed to be labile and turn over on much shorter timescales in natural ecosystems, such as POC and DOC, still have a considerable amount of SOC in 2021, up to an order of magnitude larger than the rSOC pool. While the authors mention the potential effect of the absence of biological activity in the bare follow treatments on the preservation of SOC in the conclusion, it is not clear to which extent this affects the results. For example, how useful are the results obtained from a bare-fallow experiment to parameterize models that simulate SOC cycling in vegetated soils? Would the rSOC fraction also be stable there? And would other fractions (such as POC), decline less? These are questions that cannot be answered based on the results from this data set derived from the control treatments only.
While the authors frame their results in terms of their validity for conventional SOC models containing an ‘inert’ SOC pool, it would be interesting to know how the results can be used to improve the parameterization of more recently developed SOC models that lack such a pool, and rather divide SOC into POC, MAOC and microbes, the biomass of the latter driving simulated SOC dynamics. These models generally assume that MAOC is a homogenous pool in which all C turns over at the same rate. However, the presented results contribute to an increasing amount of literature showing that ‘stabilised SOC’ is also diverse and has a wide range of turnover times (e.g., https://doi.org/10.5194/bg-18-1241-2021, https://doi.org/10.5194/bg-20-3151-2023).
Specific feedback
Title: perhaps add the word ‘experiment’ or ‘field trial’ at the end of the title?
L31: ‘[…] many SOC models exist’ would benefit from a reference, for example for one of the recent reviews on SOC models such as https://doi.org/10.1029/2023JG007436, https://doi.org/10.1029/2023JG007436 or https://doi.org/10.1016/j.soilbio.2024.109535, or some older reviews such as doi.org/10.1088/1748-9326/10/12/123004.
L36: Please explain what the AMG model is
L41: models using ‘humified organic matter’ pools have received considerable criticism over the past decade, see, for example, https://doi.org/10.1038/nature10386, https://doi.org/10.1038/nature16069 and many others. Please justify why you chose this model to link measurable C pools to, instead of more recently developed models (see review articles suggested above).
L45: please explain what the ‘splitting coefficients’ are
L55-56: in case it would appear that this fraction is indeed stable on the centennial, but not the multi-millennial scale as assumed in RothC, it would be useful to provide a recommendation (here or elsewhere in the manuscript) about the residence modelers should use for this pool (instead of the 50,000 years assumed in RothC). Or, in case this would not be possible, a more in-depth discussion about whether the turnover rate of 50,000 year in RothC is appropriate.
L67-71: any ideas on why this large difference between the French and European soils? It would be good to provide a more extensive explanation of why this difference is relevant to justify your study, for readers not familiar with the partysoc method.
L79: please mention down to which depth the topsoil samples were collected
L121: please explain in more detail how the S+A fraction was ‘retrieved with distilled water’
L138-142: more information is needed on how the statistical analyses were performed: which methods were used, and were all necessary assumptions tested?
L142: please provide the appropriate citation for the R software
L145 - 146: please explain how the C yield was measured, and how it is possible that yields up to 230% were measured. If the yield was determined as the difference in C content between the sample before and after fractionation, this requires more explanation
Fig. 2: Please provide a label for the y-axis. Please show the individual data points, in addition to the box plots. Please indicate in the methods section how significant differences were determined.
Fig. 3: Please provide a label for the y-axis
L182 (Discussion section): it would be good if a comparison with results from other studies fractionating samples collected at different points in time at long-term field trials would be made. Or if these do not exist, that this would be mentioned.
L184: ‘exhibited very limited variation’: please add that this limited variation was only in the OC% and C:N ratio, as other characteristics were not measured.
L199: ‘POC persisted’: this suggests that the composition of POC did not change (only initial POC was lost, no additional POC was created through microbial processes). As this was not measured, please rephrase.
L213: please define ‘geological coal’
L220: why specifically during the war?
L220-222: while this can indeed explain the increase in the OC% of POC in 1962, how could the (rapid) decrease in OC% towards 2021 be explained, given the long persistence of coal in soils?
L227: please explain what you mean by the ‘nature of POM’, and how this could explain the high C yields
L235: ‘DOC was rather persistent’: given the operational definition of DOC in the fractionation scheme, I would think that rather the second part of this sentence is applicable (that DOC is continuously supplied from other SOC fractions).
L247 – 250: this is a very important point, and it is not clear to me why this is only made in the conclusion. As the absence of vegetation on the studied plots will undoubtedly have an effect of the results, I would encourage the authors to discuss this in the discussion section more extensively, together with a discussion on how applicable the results are for the persistence of rSOC in soils that have vegetation.
Technical feedback
Throughout the manuscript, subscripts and superscripts are not displayed correctly
L29: ‘pathed’ => ‘paved’
L40 + L106: consists in => consists of
L125: 4g were weighted => 4g was weighed
L125 + L130: 100 mL were added => 100 mL was added
L168: unplanned => unexpected
L185: ‘temporal, stability’ => comma should not be there
Table A1 and A2: please move the caption to above the tables
Citation: https://doi.org/10.5194/egusphere-2026-1028-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 301 | 131 | 29 | 461 | 34 | 23 |
- HTML: 301
- PDF: 131
- XML: 29
- Total: 461
- BibTeX: 34
- EndNote: 23
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This study aims to explore the stability of various SOC fractionation using a centennial-scale resampling approach on a bare fallow site to. Research that tests methodological reliability like this is instructive for both our understanding of carbon cycle processes and the optimization of carbon models. However, the core conclusions remain open to question:
Together, the current evidence is not sufficient to support their core conclusions. The detailed comments are as follows:
Title：I suggest explicitly stating “centennial-scale” rather than “long-term.”
L27：I recommend that the authors insert a paragraph break, so that the meaning of the fractionation is introduced first, followed by a description of the specific techniques used.
L27-38：The section opens with the assertion that physical fractionation is more reliable than chemical methods. However, the majority of the paragraph is devoted to thermal analysis techniques. This structural imbalance distracts from the core message and may confuse readers about the primary focus of the section.
L39-45：I am curious whether the five Zimmermann’s fractions correspond one-to-one with the RothC pools. The statement in L45-46 seems to imply such a strict correspondence. However, the BIO pool in RothC is difficult to characterize using the Zimmermann fractions. It may be more closely related to MBC instead.
L74-82：How were the soil samples preserved? For a centennial-scale experiment, preservation conditions are critical to the reliability of the measurements.
L104：Does ultrasonic dispersion carry a risk of aggregate disruption, potentially leading to an underestimation of the S+C fraction?
L138-142：The elemental analyzer measures the carbon content per unit mass of each fraction, but the mass of each fraction differs. Has the carbon content been normalized to that per unit mass of air-dried soil?
L146：Major concern! The mass recovery rates in 1962 and 2021 are anomalous (in some cases exceeding 200%). This seriously undermines my confidence in the reliability of the results. This means that the authors started with 30 g of air-dried soil but ended up with 60 g of recovered soil fractions? This is physically impossible and difficult to understand.
Based on the mean mass recovery (98.16% to 100.05%, L145), I suspect that only a few samples have anomalous recoveries. I respect the authors’ transparency in reporting all data, but I recommend that they conduct a sensitivity analysis by excluding the samples with excessively high mass recoveries to test the robustness of results. In addition, reporting the carbon recovery rates is also necessary.
L175-180：The authors rely solely on visual observation to infer the source of POM and temporal changes in black carbon content. Such subjective analysis is inappropriate for presentation as formal results. Strictly, carbon source should be based on quantitative methods such as biomarker or molecular fingerprinting (e.g., 13C-NMR, Py-GC/MS). While I understand that the authors may lack the resources for such analyses, presenting subjective visual observations as evidence is scientifically unrigorous.
Discussion: The discussion would benefit from explicitly linking the findings back to the RothC model. If rSOC represents only a portion of stable SOC, then using rSOC as a direct proxy for IOM pool in RothC may systematically underestimate the soil carbon sequestration potential.
Figure 2：Please clarify the definition of sensitive and resistant SOC.