the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
How does nitrogen control soil organic matter turnover and composition? – Theory and model
Abstract. Nitrogen (N) enrichment triggers diverse responses of different soil organic carbon (SOC) pools, but a coherent mechanism to explain them is still lacking. To address this, we formulated dynamic soil CN models integrating several hypothesized N-induced decomposer responses (irrespective of plant responses), i.e., decomposition retardation under increasing N excess and stimulation under decreasing N-limitation, N-responsive microbial turnover and carbon use efficiency (CUE), and a priming effect induced by changing microbial biomass. To evaluate the relevance of each response on SOC turnover, they were incrementally combined into multiple model variants, and systematically tested against diverse observations from meta-analyses of N addition experiments and SOC fraction data from forests spanning wide environmental gradients.
Our results support the idea that N directly controls the response of multiple C pools via changing decomposition and microbial physiology. Under N addition, only the model variants that incorporated both the responses of 1) decomposition retardation with increasing N-excess and 2) decomposition stimulation with decreasing N limitation were able to reproduce the common observation of a greater increase of surface organic horizon (LFH) relative to topsoil SOC, and of particulate organic carbon (POC) relative to mineral-associated carbon (MAOC). In addition, cold and warm forests respectively experienced more decomposition retardation and stimulation under N addition. Furthermore, incorporating N-responsive microbial turnover and CUE helped reproduce microbial biomass reduction, and the latter was also critical for microbial biomass C:N homeostasis, which in turn constrained the estimation of N-limitation and excess.
Synthesizing the model findings and literature, we propose that N addition accelerates the decomposition of N-limited detritus, which supplies C to intermediate processed pools (i.e., light fraction C), and retards the decomposition of processed organic matter with lower C:N ratios (both light fraction C and MAOC). This explains the large light fraction C accumulation under N addition or contemporary N deposition in temperate forests. Collectively, our model experiment provided robust mechanistic insights on soil N-C interaction, and challenged the common model assumption of plant being the primary respondent to N. We recommend our simple model for further testing and ecological applications.
Competing interests: Frank Hagedorn is on the editorial board of Biogeosciences.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(2217 KB) - Metadata XML
-
Supplement
(1307 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-1022', Anonymous Referee #1, 14 May 2025
In this manuscript, Yeung et al. develop a new framework and model for the effect of N addition on soil organic matter cycling, accounting for both enhanced decomposition of N-poor substrates/soils and slower decomposition in N-rich soils and MAOM. The theory is tested using a model experiment, and the results compared to a meta-analysis of experimental responses to N addition.
Overall, I think the manuscript will be an excellent contribution to the literature. The model experiment approach is compelling evidence for the hypothesized framework. The writing is very clear and the model is well-explained.
Major comment:
How can the decomposition retardation framework (equations 3 to 5) can apply to both SOC2 (lignin-containing materials) and protein-rich MAOC? From the equations, it appears the effect is driven by the difference of the C:N of the pool compared to TER, but it seems like the C:N of SOC2 would nearly always be higher than TER. How would this simulate the observed inhibition of lignin-degrading enzymes?
Furthermore, the framework shown in figure 1 implies that decomposition lignin-rich but N-poor substrates like deadwood increase with N addition.
Minor comments:
Line 159: “parameters” rather than “parameter”
Line 169: Suggest “turns over” instead of “turnovers”
Line 187: suggest adding “see section 2.6”, as the meta-analysis approach has not been introduced at this point in the manuscript.
Citation: https://doi.org/10.5194/egusphere-2025-1022-RC1 - AC2: 'Reply on RC1', Chun Chung Yeung, 02 Jul 2025
-
CC1: 'Comment on egusphere-2025-1022', Marijn Van de Broek, 23 May 2025
Dear Authors,
I have read the manuscript ‘How does nitrogen control soil organic matter turnover and composition? – Theory and model’ with much interest. Unravelling the different processes affecting soil C – N interactions is an important topic, as is the correct incorporation of these processes in mechanistic soil organic matter models. With my contribution to the open discussion, I would like to focus on a few modelling aspects of the study. I hope my feedback helps the authors to clarify these aspects in the manuscript.
My main feedback relates to the choice of the model (CENTURY). The structure of this model is based on the humification theory, which has been criticised in the literature (Lehmann and Kleber, 2015 (doi.org/10.1038/nature16069); Kleber and Lehmann, 2019 (doi.org/10.2134/jeq2019.01.0036)) and is considered outdated. Although the names of the pools have been changed to measurable pools, as was done in previous studies, the structure is different from the general understanding of SOM cycling (e.g., there is no DOC pool, there is a direct flux of POC to MAOC, although in reality POC is depolymerised to DOC first, which can then either be adsorbed on minerals, taken up by microbes or leached from the soil, and the amount of soil microbes does not affect the rate of decomposition (although this is implemented as a scenario in the study)). However, the authors consistently refer to this model as ‘process-based’. Over the past two decades, multiple models simulating coupled soil C – N cycles using measurable pools, while explicitly simulating established mechanisms which are addressed in the present study (e.g., adsorption and desorption of DOC on/from soil minerals, the effect of microbial biomass (or enzymes) on rates of polymerisation and decomposition, microbial CUE, etc.). Therefore, I would suggest the authors to better justify their choice for this model, over other, more mechanistic, SOM models, as these explicitly incorporate many of the simulated mechanisms related to soil microbes and MAOC discussed in the present manuscript.
Related to this, in L462, the authors cite studies using models which are very similar (Chen et al., 2019; CASA CNP in Eastman et al., 2024), or more mechanistic (MIMICS in Eastman et al., 2024) to the model they used, and refer to these as ‘process-deficient’. How do the authors reconcile this with referring to CENTURY as a mechanistic model? In the same sentence, they refer to models containing multiple mechanisms as containing ‘many uncertain parameters and hence [it is] difficult to verify the relevance of each process’. However, in their study the authors use mainly default parameters for the CENTURY and FORCENT models. I would encourage the authors to explain why the parameter values in the latter would be less uncertain compared to those in state-of-the-art mechanistic SOM models.
In their study, the authors compare simulated effects of N addition to Swiss forest soils with the results from a global meta-analysis. I would encourage the authors to better (graphically) represent the differences in environmental characteristics between the Swiss sites and the sites from the meta-analyses. Related to this, the authors justify the comparison to results from global meta-analyses by stating that ‘most experiments were conducted in temperate forests’ (L330). Which portion is ‘most’? It would be good to justify why not only these experiments from temperate forests were used in the comparison, instead of also using experiments from other ecosystems.
Specific feedback
L31-32 and L637-638: You state that your result challenge the ‘common assumption that plant is the primary respondent to N’. This is a strong statement, as you (from what I understood in the manuscript) did not simulate the response of plants to N addition, or interactions between the response of plants and soil microbes. Therefore, it makes sense that without simulating the response of plants, the response of soils will be large. However, as the response of plants wasn’t quantified, I suggest to revise this statement.
L52: you state that ‘Efforts to synthesize the diverse responses into a coherent theory and model are lacking’, although many modelling studies have studied this effect, of which you cite a few. Therefore, I encourage the authors to modify this statement. In addition, as this is a modelling study, it would have been useful to summarise the outcomes of these previous modelling studies with respect to the response of soil organic matter characteristics to N addition.
L166: Was the turnover rate parameter for SOC3 adjusted based on simulations of the total SOC stock, or were measurements of MAOC stocks available? It would also be useful to mention the new value for the turnover rate parameter for SOC3 here. Is this value in line with reported turnover times of MAOC in forest topsoils?
L187: It would be good to provide references for these meta-analyses.
Section 2.3: I appreciate the authors testing many N responses. I would encourage the authors to provide graphs that show how the proposed variables and rate modifiers vary in response to the independent variables, as this is difficult to understand from the equations alone. In addition, it would be good to mention if the proposed equations are based on literature, or if the authors formulated these equations themselves.
L244: how is the ‘monthly priming effect’ defined?
L254-255: I suggest reconsidering the subscript ‘retard’
L259: It would be good to provide citations for these ‘widespread experiments’
L287 and L301: the correlation coefficient ‘Pearson r’ is not a measure for model performance, but for the proportion of explained variation. I would suggest reporting an error measure such as RMSE instead.
Section 2.5: it would be good to mention here down to which depth the simulations were performed
L307: ‘[…] until the pools were at steady-state’: for a better evaluation of the model performance by the reader, it would be good to report the simulated turnover times of the model pools, and how C was distributed between SOC1, SOC2 and SOC3.
L318-327: as the data was extracted from global meta-analyses, it would be good to report, for example, from which ecosystems these data were collected. In addition to table 2, the authors can consider a graphical representation of the environmental characteristics of the sites from the meta-analyses, overlain by the characteristics for the Swiss forests. This information is essential for the reader to evaluate the comparison that is made with the Swiss forest sites.
Related to this, on L575 the authors state the ‘environmental gradient coverage [of the Swiss sites] differs *slightly*’ from the sites in the meta-analyses, while, for example, MAT for the Swiss sites ranges from 6.5 to 9.4 °C, while this is between 1.0 and 21 °C for the sites from the meta-analyses. This statement thus seems misleading, and a better representation of these differences seems necessary to inform the reader, as the comparison between the simulation results obtained from the Swiss sites and those from the meta-analyses is at the core of this study.
L354-358: It seems that the base model was not able to accurately simulate total SOC stocks for the Swiss forest sites (Fig. 3b). The authors justify this by stating that ‘a close quantitative match of C stocks may not actually be desirable […]’. However, I encourage the authors to better justify why they thrust results of a model that is not able to simulate total SOC stocks accurately, and to better explain why this was the case. For example, why were the model parameters not calibrated to match the conditions in Swiss forests? To evaluate the model results, an even more important aspect of the model outcomes is the simulated distribution of SOC and N over the different model pools, and the simulated turnover times of these pools (e.g., which portion of simulated SOC was present in SOC1, SOC2 and SOC3). I would encourage the authors to also report this information.
L368: It would be good to report the value for the ‘low SOC3 parameter’ here
Fig. 5a: while the regression line looks fine, a large portion of the simulated data is overestimated (y-values between 13-16), while another portion is underestimated (y-values between 7-9). This does not seem in line with the statement of ‘the models capturing MAOC:N reasonably’. I would suggest the authors to revise this statement.
L461-465: instead of dismissing previous modelling studies as ‘process-deficient’ or ‘containing many uncertain parameters’, it would have been worthwhile to summarize their findings in the introduction as a background to the modelling study that is presented here.
L485: ‘enhanced SOC3 turnover’: I would suggest to report this turnover rate and evaluate if this is in line with previously reported turnover rates of MAOC in forest soils.
L509: what is meant by ‘simplistic’ decomposition?
Tables and Figures
Fig. 1: It was not clear to me what the y-axis label ‘Labile C:N’ means. Can this be clarified?
Table 1: In the second row under (2.), should it be ‘increasing N limitation?’, instead of decreasing?
Fig. 3: a unit needs to be reported for the RMSE
Fig. 7: it would be good to provide more information about this figure in the caption.
Citation: https://doi.org/10.5194/egusphere-2025-1022-CC1 - AC3: 'Reply on CC1', Chun Chung Yeung, 02 Jul 2025
-
RC2: 'Comment on egusphere-2025-1022', Anonymous Referee #2, 01 Jun 2025
egusphere-2025-1022_Review comments
This study makes a significant contribution by demonstrating that direct microbial responses to N stoichiometry can explain soil organic carbon (SOC) pool dynamics across environmental gradients, thereby challenging plant-centric paradigms. The incremental model design and multi-scale validation provide a rigorous framework for understanding N-C interactions. However, limitations in microbial representation, parameter uncertainty, and plant-microbe feedbacks highlight opportunities for future refinement. Overall, the work advances our mechanistic understanding of soil N-C coupling and offers a valuable foundation for integrating decomposer ecology into global change models. There are some concerns that the authors should consider for further improvement of the manuscript's current quality. These review comments challenge the model’s assumptions in this study, highlight gaps in mechanistic representation, and invite further validation across scales and ecosystems.
For my overall review decision, I would like to suggest this article for a major revision for further consideration for acceptance by the journal.
- In the Introduction part, what is the “state-of-the-art” soil model? Are they really “state-of-the-art”? How to define such a confusing concept? Additionally, the authors would be better to give some examples in detail to further explain such “state-of-the-art”.
- In this article, the authors oversimplified the microbial dynamics. The model aggregates microbes into a single pool per soil layer, ignoring functional guilds (e.g., fungi vs. bacteria) that differ in N-use efficiency and substrate preferences. For instance, ectomycorrhizal fungi dominate lignin decomposition in boreal forests, but their response to N (e.g., suppression under N-excess) is not explicitly modeled. Additionally, it is about the TER (i.e., static TER). TER is calculated as microbial biomass C: N divided by maximum CUE, but this may oversimplify microbial adaptability. Recent studies have shown that TER can vary in response to nutrient stress and the dynamics of climate change, which could alter predictions of N-excess and limitation thresholds. I am curious about the model test scenarios where TER adapts to chronic N addition, and how this would affect estimates of N-excess vs. N-limitation?
- The MAOC turnover and pH effect are uncertain in the model discussion. For example, MAOC is assumed to be a homogeneous "slow-turnover" pool, yet real-world MAOC contains labile sub-fractions, which could explain the model’s short-term MAOC response lag. Additionally, The base model includes pH retardation on decomposition but assumes uniform pH across soil layers, contradicting evidence that FH horizons are more acidic than mineral soil. This may overestimate LFH accumulation in acidic forest soils.
- While the model tests MAT and MAP effects, it does not explicitly link temperature to microbial enzyme kinetics or N mineralization rates. For example, in warm forests, N addition may accelerate the mineralization of labile C through temperature-microbial interactions, which are not captured here. Contemporary N deposition data are assumed to be representative of long-term trends; however, historical peaks (e.g., the 1980s in Europe) may have legacy effects on SOC that are not reflected in the model. Additionally, while the Swiss forest dataset is robust, testing the model in other biomes (e.g., boreal, tropical forests) would validate its generality. For example, N-limitation is more pronounced in tropical soils, potentially amplifying decomposition stimulation effects on high C:N litter. So, in tropical forests with high C:N litter and strong N limitation, would decomposition stimulation under N addition be more pronounced than simulated?
- By excluding plant responses (e.g., root exudation, litter quality changes), the model may underestimate N effects in systems where plants dominate C input shifts. For instance, N-induced shifts from fine roots to aboveground litter could alter SOC pool dynamics, a pathway not explored here. Would incorporating plant-derived changes in litter quality (e.g., lower lignin: C ratios) alter the predicted dominance of microbial vs. plant-driven SOC changes? Fine root decomposition contributes to POC and MAOC, but the model treats all litter inputs uniformly. How does the lack of root-soil carbon pathway distinction affect predictions of N-induced POC accumulation?
- Nitrogen Excess vs. Acidification??? N addition and acidification often co-occur, but the model attributes decomposition retardation to N-excess alone. How were the effects of N-induced acidification disentangled from stoichiometric imbalances, especially in acidic Swiss forests (pH range is not narrow, instead, it is large)?
- The model evaluates N-addition effects over 5–15 years, aligning with most experimental durations, but meta-analyses show divergent long-term trends (e.g., MAOC accumulation). Given that microbial necromass and mineral-associated C turnover occur over decades to centuries, how does the model validate its precision in predicting transient vs. asymptotic SOC responses, especially for slow-turning pools like MAOC, and are there plans to incorporate time-dependent microbial adaptation (e.g., enzyme induction/suppression) to improve long-term predictability? Can the model’s long-term MAOC predictions be validated using centennial-scale soil carbon datasets?
Citation: https://doi.org/10.5194/egusphere-2025-1022-RC2 - AC1: 'Reply on RC2', Chun Chung Yeung, 02 Jul 2025
Data sets
Model input and validation datasets Chung C. Yeung, Harald Bugmann, Frank Hagedorn, Margaux Moreno Duborgel, and Olalla Díaz-Yáñez https://doi.org/10.5281/zenodo.14879678
Model code and software
Model source code and analysis code Chung C. Yeung, Harald Bugmann, Frank Hagedorn, Margaux Moreno Duborgel, and Olalla Díaz-Yáñez https://doi.org/10.5281/zenodo.14879678
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
405 | 136 | 21 | 562 | 13 | 11 | 34 |
- HTML: 405
- PDF: 136
- XML: 21
- Total: 562
- Supplement: 13
- BibTeX: 11
- EndNote: 34
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1