the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Radiocarbon analysis reveals underestimation of soil organic carbon persistence in new-generation soil models
Abstract. Reflecting recent advances in our understanding of soil organic carbon (SOC) turnover and persistence, a new generation of models increasingly makes the distinction between the more labile soil particulate organic matter (POM) and the more persistent mineral-associated organic matter (MAOM). Unlike the typically poorly defined conceptual pools of traditional SOC models, the POM and MAOM pools can be directly measured for their carbon content and isotopic composition, allowing for pool-specific data assimilation. However, the new-generation models' predictions of POM and MAOM dynamics have not yet been validated with pool-specific carbon and 14C observations. In this study, we evaluate 5 influential and actively developed new-generation models (CORPSE, Millennial, MEND, MIMICS, SOMic) with pool-specific and bulk soil 14C measurements of 77 mineral topsoil profiles in the International Soil Radiocarbon Database (ISRaD). We find that all 5 models consistently overestimate the 14C content (Δ14C) of POM by 67 ‰ on average, and 3 out of the 5 models also strongly overestimate the Δ14C of MAOM by 74 ‰ on average, indicating that the models generally overestimate the turnover rates of SOC and do not adequately represent the long-term stabilization of carbon in soils. These results call for more widespread usage of pool-specific carbon and 14C measurements for parameter calibration, and may even suggest that some new-generation models might need to restructure their simulated pools (e.g., by adding inert pools to POM and MAOM) in order to accurately reproduce SOC dynamics.
- Preprint
(2741 KB) - Metadata XML
-
Supplement
(9279 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
CEC1: 'Comment on gmd-2023-242', Juan Antonio Añel, 26 Jan 2024
Dear authors,
Unfortunately, after checking your manuscript, it has come to our attention that it does not comply with our "Code and Data Policy".
https://www.geoscientific-model-development.net/policies/code_and_data_policy.html
You have archived your code on GitHub. However, GitHub is not a suitable repository for scientific publication. GitHub itself instructs authors to use other alternatives for long-term archival and publishing, such as Zenodo. Therefore, please, publish your code in one of the appropriate repositories, and reply to this comment with the relevant information (links and DOIs) as soon as possible, as it should be available before the Discussions stage.Note that if you do not fix this problem, we will have to reject your manuscript for publication in our journal. I should note that, actually, your manuscript should not have been accepted in Discussions, given this lack of compliance with our policy. Therefore, the current situation with your manuscript is irregular.
Also, you must include in a potentially reviewed version of your manuscript the modified 'Code and Data Availability' section, the DOIs of the code (and another DOI for the dataset if necessary).
Also, you include several appendices (A1, A2, etc.) that actually contain information on the code. Such information must be part of the "Code Availability" section, and you should move the contents of such appendices to it
Juan A. Añel
Geosci. Model Dev. Executive EditorCitation: https://doi.org/10.5194/gmd-2023-242-CEC1 -
AC1: 'Reply on CEC1', Alexander Brunmayr, 27 Jan 2024
Dear Dr. Añel,
Apologies for the delay in producing a DOI for our code. Here is the DOI for the Zenodo archive of our code:
http://doi.org/10.5281/zenodo.10575139
I do not yet have the DOIs for all of the evaluated models' source codes. I am contacting the model developers to ask them if they could either publish their code on Zenodo themselves or provide the relevant information (list of creators and contributors, choice of license, funding sources, etc.) for me to publish their code on Zenodo. I will then update the "Code and Data Availability" section with the DOIs once they are ready.
On your last point, I will move all references and links to the source codes from Appendix A to the "Code and Data Availability" section. However, Appendix A also goes into the details of how and why we adapted the original source codes for our purposes, and explains which exact equations, parameter values, and spin-up methods we used for each model. Is this sort of information appropriate and relevant for the "Code and Data Availability" section? Should we move the entire text of Appendix A into "Code and Data Availability"?
Thank you,
AlexanderAlexander S. Brunmayr (Author comment on behalf of all Co-Authors)
Citation: https://doi.org/10.5194/gmd-2023-242-AC1
-
AC1: 'Reply on CEC1', Alexander Brunmayr, 27 Jan 2024
-
RC1: 'Comment on gmd-2023-242', Jeffrey Beem Miller, 02 Feb 2024
New generation soil models include mechanistic pools, but lack validation. The authors of this work sought to address this research gap by evaluating 5 new generation soil C models using C and 14C data from SOM density fractions obtained from the International Soil Radiocarbon Database (ISRaD), focusing on topsoil. They found that the ∆14C of particulate organic matter (POM) pools were consistently overestimated (i.e., enriched) in the models relative to the observations, and the majority of the models overestimated ∆14C of mineral associated soil organic matter (MAOM) as well. The authors conclude that models could be improved through additional assimilation of measured 14C and C fraction data, which in turn may inform changes to model structures that would better fit available observations. As a specific recommendation, the authors point to the inclusion of an “inert” pool in the new generation models (a typical feature of 1st generation models such as CENTURY and RothC) to improve the fit to observed 14C data.
Overall, I found this manuscript to be a well-written and timely contribution to the field. However, there are a few technical issues with respect to the interpretation of the 14C data and what I feel to be some missed opportunities for comparing and contrasting the model results. I also have a concern about a fundamental premise of the work, i.e., that density fractions can be considered homogeneous pools. Perhaps this is more of a semantic issue, but I believe it warrants clarification.
I noted two key issues with the interpretation of the 14C data. First is the influence of the year of sampling, which is particularly relevant for assessing the relationship of ∆14C values to environmental variables with a linear regression approach. This source of potential bias can be addressed by normalizing ∆14C to a common year of sampling, as performed in Shi et al. (2020) and Heckman et al. (2022) (references listed below). The ISRaD R package provides a function for performing this normalization (ISRaD.extra.norm14c_year). More details on the specific concerns are outlined in the specific comment on lines 201-202. Second, the extremely strong dependence of 14C with depth in soils can lead to serious issues when drawing conclusions across studies with differing depth increments. Did you predict specific depths from the models? Did you spline the depth profiles from ISRaD (cf. Bishop et al., 1999; Malone et al., 2009; Sierra et al., 2024)? More information is needed here, and perhaps some adjustments to your approach.
Another area in which I think the study could be strengthened is further exploration of the nuances of the fraction data and the pool-specific model results. For example, I would be interested to know why the occluded particulate organic matter (oPOM) fraction was not explored further (which in some soils contains older C than the heavy fraction). Or, given the current understanding that the heavy fraction is also likely a mixture of organic matter cycling across a range of time scales (and via distinct mechanisms), does the comparison of the observed data with more explicit pools of the models shed light on the environmental conditions under which MAOM is cycling relatively faster or slower (Heckman et al., 2018; Stoner, 2023)? There is some discussion of how the turn-over times assigned to the micro-aggregate and mineral adsorbed pools of the Millenial model are too fast (lines 235-236). Are there examples from the other models that can be compared and contrasted here? Can you draw some conclusions about which (putatively) measureable pools should be targeted to improve future models?
The authors identify the lack of a very slow pool in the models as a common issue across all of the models, which is a strong finding. However, the suggestion of simply adding an “inert” pool to compensate is not a novel concept, and in my view, does not help to move the field forward. Case in point, the slowest pool in the original RothC model (Jenkinson and Rayner, 1977) was “chemically stabilized soil organic matter” with a turnover of 1980 y. The “inert” pool was only introduced as a “deus ex machina” and less “elegant” solution (to quote the authors) for improving the fit of the model to observed 14C (Jenkinson, 1990). Furthermore, the suggestion that this inert material is equivalent to pyrogenic C is not convincing. Unfortunately, I do not think this is likely to be a useful solution either, given the known heterogeneity of pyrogenic C, which makes it ill-suited as a functional pool of SOM (unless further subdivided). The discussion (and analysis) could benefit from further discussion of how pyrogenic C may be incorporated into functional pools such as aggregates, or mineral-association, as well as the role of other potentially long-lived soil C pools.
Finally, the issue of whether empirical fractions can be considered modelable pools is at the heart of this work. Yet the terms “POM” and “MAOM” are used interchangeably for both empirical fractions and model pools throughout the paper, despite that fact that your results undermine this equivalence. I recognize that nomenclature is always a challenge, however, I think it needs to be made clearer that this is a hypothesis that is being tested, rather than an accepted fact. In light of your results, I would argue that a key addition to (or modification of) the list of likely reasons why the new-generation models underperform with respect to 14C (lines 225-229) is precisely the underlying heterogeneity of the measured fractions: i.e., the fact that these putative “measureable pools” are in fact a mixture of different pools of soil organic matter cycling at distinct rates, which in turn, are determined by different persistence mechanisms. Indeed, this paradigm is referenced briefly later in the discussion, especially with respect to the specific point about the role of pyrogenic C. Put another way, perhaps you could try to quantify what the benefits of density fractionation are for understanding soil organic matter dynamics, in light of the limitations that have been demonstrated here.
Is a key take-away from this work that we should continue performing density fraction, but simply take more care to separate charcoal from the light fractions? My understanding of the results is that this is not the case, and it seems like there is the potential here to provide more nuanced (and needed!) advice on how to move forward with modeling the measurable.
Line specific notes:
- Ln 42-43: misleading, not just pedogenic oxides that form mineral-organic associations (Kleber et al., 2015). Clarify calcium bridging as a distinct mechanism in non-acidic soils (distinct from OM adsorption via surface charges of pedogenic oxides).
- Ln 44: define residence time, cf. (Sierra et al., 2017)
- Ln: 49: the Wagai reference here explicitly discusses the importance of mineral coating for the occluded light fraction. Most POM does not have mineral coatings.
- Ln 51: can you elaborate on how this micro aggregate protection is related to occlusion or distinct from adsorption mechanisms?
- Ln 52-54: This is a bit of an overstatement. Yes, there is the potential to measure the model pools, and it is true these models were designed, by and large, for the pools to be measureable. However, as this paper clearly identifies, much work remains to be done to make adequate empirical measurements of these new generation model pools.
- Ln 72: Could also cite Metzler et al. (2020), who developed a mathematical approach for calculating ∆14C from any compartmental model, i.e., expressible as system of ODEs
- Ln 97: I understand why you have made this nomenclature decision, but I am uncomfortable with the underlying assumption that these fractions are equal to the pools. A suggestion could be to add subscripts indicating measured vs modeled? Or please be clear that this is a hypothesis that is being tested here.
- Ln 102: Please elaborate more on how you dealt with depth here (see notes above).
- Ln 103: ISRaD contains substantially more density fraction 14C data than this. Can you clarify how you made your selections? Additionally, contrary to your statement in line 109, there are at least two studies with density fraction data from permafrost soils (O’Donnell et al., 2011; Gentsch et al., 2018).
- Ln 158: Do these 14C data account for potential pre-aging in vegetation (Gaudinski et al., 2000; Joslin et al., 2006; Herrera-Ramírez et al., 2020)?
- Ln 179: This is a bit confusing as written. He et al. (2016) found that ESMs consistently underestimated the age of soil C, which is also what you are suggesting here. But in terms of overestimating ∆14C, that has a rather different interpretation when you are talking about pre-bomb ∆14C values below 0, as the results from He et al. show, versus modern era ∆14C values above 0, as you are showing. In the case of He et al., this indicated models were underestimating soil C ages by ~ hundreds of years. In your case it is much more ambiguous, and may only indicate differences of tens of years. Perhaps you could reword this to clarify what you mean.
- Ln 192: Why was this site selected? Is it representative?
- Ln 195: perhaps refer to the appendix?
- Ln 201-202: This explanation needs some more interpretation (and is probably more suited the discussion). Given the large amount of change observed in atmospheric ∆14C over the period 1997-2015 (103 to 14 per mil) (Graven et al., 2017), these trends are ambiguous. For example, it could be just as likely that POM ∆14C would decrease with increasing temperature (as it does in some of the models) as the pool is cycling faster and is therefore closer to the declining atmospheric signal. The predicted trend should be a function of both the partitioning of C into the pool (its size) and its turnover rate.
- Ln 207-208: Clarify observed vs. modeled?
- Ln 217: Again, careful with the comparison to He et al. 2016. Overestimation of ∆14C means something very different in the context of that study versus this one (see note at line 179)
- Ln 219: Perhaps “partitioning” instead of “repartition”?
- Ln 230: Please elaborate more on the effectiveness of this distinction.
- Ln 240: Perhaps “putatively” measureable pools?
- Ln 249: “practically inert” is perhaps misleading here. The intent is clear, but many studies have confirmed that there is no “inert” soil C, only soil C that is inaccessible or either stoichiometrically or energetically unfavorable for microbes to consume.
- Ln 252: “extreme longevity” this is not technically correct, as pyrogenic C is known to exhibit a wide range of 14C values with inferred mean residence times from decadal to centennial timescales (cf. Fig. 1, Schmidt et al., 2011)
- Ln 254: consider citing (Grant et al., 2023)
- Ln 255: And indeed, this was due to these models being informed by soil 14C measurements, e.g., Jenkinson and Rayner, 1977.
- Ln 258-260: Reintroducing an “inert” pool seems to be contrary to the goals of these new models of having measurable pools in the models, as outlined by Lehmann and Kleber (2015) and also creates instability issues when searching for analytical model solutions (Sierra and Mueller, 2015).
- Ln 266-268: Yes, this is a key point!
- Ln 270: I would be cautious in assuming that pyrogenic C itself is a functional pool. The bigger issue here is that it violates the premise that density fractionation yields homogenous pools.
- Ln 285-287: As mentioned previously, pre-aging of C in vegetation is not only an issue with thick O horizons.
- Ln 293: Please reword.
- Ln 371-372: This contradicts the definition of POM as supplied in the introduction.
References
Bishop, T. F. a, McBratney, a. B., and Laslett, G. M.: Modeling soil attribute depth functions with equal-area quadratic smoothing splines, Geoderma, 91, 27–45, https://doi.org/10.1016/S0016-7061(99)00003-8, 1999.
Gaudinski, J. B., Trumbore, S. E., Davidson, E. A., and Zheng, S.: Soil carbon cycling in a temperate forest: radiocarbon-based estimates of residence times, sequestration rates and partitioning of fluxes, Biogeochemistry, 51, 33–69, https://doi.org/10.1023/A:1006301010014, 2000.
Gentsch, N., Wild, B., Mikutta, R., Čapek, P., Diáková, K., Schrumpf, M., Turner, S., Minnich, C., Schaarschmidt, F., Shibistova, O., Schnecker, J., Urich, T., Gittel, A., Šantrůčková, H., Bárta, J., Lashchinskiy, N., Fuß, R., Richter, A., and Guggenberger, G.: Temperature response of permafrost soil carbon is attenuated by mineral protection, Global Change Biology, 24, 3401–3415, https://doi.org/10.1111/gcb.14316, 2018.
Grant, K. E., Hilton, R. G., and Galy, V. V.: Global patterns of radiocarbon depletion in subsoil linked to rock-derived organic carbon, Geochem. Persp. Let., 25, 36–40, https://doi.org/10.7185/geochemlet.2312, 2023.
Graven, H., Allison, C. E., Etheridge, D. M., Hammer, S., Keeling, R. F., Levin, I., Meijer, H. A. J., Rubino, M., Tans, P. P., Trudinger, C. M., Vaughn, B. H., and White, J. W. C.: Compiled records of carbon isotopes in atmospheric CO2 for historical simulations in CMIP6, Geoscientific Model Development, 10, 4405–4417, https://doi.org/10.5194/gmd-10-4405-2017, 2017.
Heckman, K., Lawrence, C. R., and Harden, J. W.: A sequential selective dissolution method to quantify storage and stability of organic carbon associated with Al and Fe hydroxide phases, Geoderma, 312, 24–35, https://doi.org/10.1016/j.geoderma.2017.09.043, 2018.
Heckman, K., Hicks Pries, C. E., Lawrence, C. R., Rasmussen, C., Crow, S. E., Hoyt, A. M., von Fromm, S. F., Shi, Z., Stoner, S., McGrath, C., Beem-Miller, J., Berhe, A. A., Blankinship, J. C., Keiluweit, M., Marín-Spiotta, E., Monroe, J. G., Plante, A. F., Schimel, J., Sierra, C. A., Thompson, A., and Wagai, R.: Beyond bulk: Density fractions explain heterogeneity in global soil carbon abundance and persistence, Global Change Biology, 28, 1178–1196, https://doi.org/10.1111/gcb.16023, 2022.
Herrera-Ramírez, D., Muhr, J., Hartmann, H., Römermann, C., Trumbore, S., and Sierra, C. A.: Probability distributions of nonstructural carbon ages and transit times provide insights into carbon allocation dynamics of mature trees, New Phytologist, 226, 1299–1311, 2020.
Jenkinson, D. S.: The turnover of organic carbon and nitrogen in soil, Philosophical Transactions: Biological Sciences, 329, 361–368, 1990.
Jenkinson, D. S. and Rayner, J. H.: THE TURNOVER OF SOIL ORGANIC MATTER IN SOME OF THE ROTHAMSTED CLASSICAL EXPERIMENTS, Soil Science, 123, 298, 1977.
Joslin, J. D., Gaudinski, J. B., Torn, M. S., Riley, W. J., and Hanson, P. J.: Fine-root turnover patterns and their relationship to root diameter and soil depth in a 14C-labeled hardwood forest, New Phytologist, 172, 523–535, https://doi.org/10.1111/j.1469-8137.2006.01847.x, 2006.
Kleber, M., Eusterhues, K., Keiluweit, M., Mikutta, C., Mikutta, R., and Nico, P. S.: Mineral-Organic Associations: Formation, Properties, and Relevance in Soil Environments, Elsevier Ltd, 1–140 pp., https://doi.org/10.1016/bs.agron.2014.10.005, 2015.
Lehmann, J. and Kleber, M.: The contentious nature of soil organic matter, Nature, 528, 60–68, https://doi.org/10.1038/nature16069, 2015.
Malone, B. P., McBratney, a B., Minasny, B., and Laslett, G. M.: Mapping continuous depth functions of soil carbon storage and available water capacity, Geoderma, 154, 138–152, https://doi.org/10.1016/j.geoderma.2009.10.007, 2009.
Metzler, H., Zhu, Q., Riley, W., Hoyt, A., Müller, M., and Sierra, C. A.: Mathematical Reconstruction of Land Carbon Models From Their Numerical Output: Computing Soil Radiocarbon From C Dynamics, Journal of Advances in Modeling Earth Systems, 12, e2019MS001776, https://doi.org/10.1029/2019MS001776, 2020.
O’donnell, J. A., Harden, J. W., McGUIRE, A. D., Kanevskiy, M. Z., Jorgenson, M. T., and Xu, X.: The effect of fire and permafrost interactions on soil carbon accumulation in an upland black spruce ecosystem of interior Alaska: implications for post-thaw carbon loss, Global Change Biology, 17, 1461–1474, https://doi.org/10.1111/j.1365-2486.2010.02358.x, 2011.
Schmidt, M. W. I., Torn, M. S., Abiven, S., Dittmar, T., Guggenberger, G., Janssens, I. a., Kleber, M., Kögel-Knabner, I., Lehmann, J., Manning, D. a. C., Nannipieri, P., Rasse, D. P., Weiner, S., and Trumbore, S. E.: Persistence of soil organic matter as an ecosystem property, Nature, 478, 49–56, https://doi.org/10.1038/nature10386, 2011.
Shi, Z., Allison, S. D., He, Y., Levine, P. A., Hoyt, A. M., Beem-Miller, J., Zhu, Q., Wieder, W. R., Trumbore, S., and Randerson, J. T.: The age distribution of global soil carbon inferred from radiocarbon measurements, Nature Geoscience, 13, 555–559, https://doi.org/10.1038/s41561-020-0596-z, 2020.
Sierra, C. A. and Mueller, M.: A general mathematical framework for representing soil organic matter dynamics, Ecological Monographs, 85, 505–524, https://doi.org/10.1890/07-1861.1, 2015.
Sierra, C. A., Müller, M., Metzler, H., Manzoni, S., and Trumbore, S. E.: The muddle of ages, turnover, transit, and residence times in the carbon cycle, Global Change Biology, 23, 1763–1773, https://doi.org/10.1111/gcb.13556, 2017.
Sierra, C. A., Ahrens, B., Bolinder, M. A., Braakhekke, M. C., von Fromm, S., Kätterer, T., Luo, Z., Parvin, N., and Wang, G.: Carbon sequestration in the subsoil and the time required to stabilize carbon for climate change mitigation, Global Change Biology, 30, e17153, https://doi.org/10.1111/gcb.17153, 2024.
Stoner, S.: Quantifying relevant timescales of soil carbon cycling through long-term modeling and novel fractionation techniques, phd, ETH Zurich, 2023.
Citation: https://doi.org/10.5194/gmd-2023-242-RC1 -
RC2: 'Comment on gmd-2023-242', Anonymous Referee #2, 18 Feb 2024
The authors compile a suite of complex “new generation” soil carbon models that have either implemented radiocarbon into their model structure or have the ability to do so. They fix errors in previous 14C implementation in 2 models and run all models with similar inputs at a set of sites where soil D14C of soil fraction data has been collected and stored in the ISRAD Database. They find that there is generally an overestimation of soil D14C in these models, indicating that they are underestimating the degree of persistence of soil carbon. Overall, this is a good contribution to the soil carbon model community, fixing inaccuracies in 2 models and highlighting that the problems in carbon turnover are not solved by just adding more complexity.
For the measured data from ISRaD, the authors should consider adding sites that have just bulk carbon density and D14C data, even if they do not have the size fraction D14C data, to test if the consideration of more sites (spatially and temporally) changes the result.
While the figures are generally well-designed and easy to understand, I think it would be interesting to see the spatial patterns of how well these models are doing. I think one more figure of the spatial pattern of bulk soil carbon and bulk D14C RMSEs for specific sites may be helpful in this. Another metric that should be considered is the model RMSE of the difference between POM and MAOM D14C.
Line Comments:
87: “used” or “use” instead of “will use”
137-138: Details of the spinup of these models should be added to this section instead of the Appendix because the spinup time can significantly affect the D14C values. Also concerning the spinup, details in the Appendix are needed on the percentage of sites in each model that reached equilibrium “solver” values vs. the base 4000-5000 year periods.
192: Figure 6 makes me wonder if plotting the difference between the D14C values of MOAM and POM for all sites might be beneficial in model comparison. This figure is making it seem like the MIMICS model might be getting the bulk D14C right for the wrong reasons, but it is only one site. You could add this metric to Figure 5 and add more discussion of this difference.
505: Details are needed on how much this difference in the original MIMICS D14C implementation affects bulk, POM and MAOM D14C values in the new implementation. Something similar to line 491 for the SOMic model should be fine.
Citation: https://doi.org/10.5194/gmd-2023-242-RC2
Model code and software
evaluate-SOC-models Alexander S. Brunmayr https://github.com/asb219/evaluate-SOC-models
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
270 | 89 | 22 | 381 | 25 | 16 | 12 |
- HTML: 270
- PDF: 89
- XML: 22
- Total: 381
- Supplement: 25
- BibTeX: 16
- EndNote: 12
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1