the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 16 Jul 2026
A process-based framework for national-scale estimation of agricultural soil N₂O emissions under variable climate and management
Abstract. Agricultural soils are the dominant source of anthropogenic N2O emissions, yet their high spatial and temporal heterogeneity provides a major challenge for accurately quantifying emissions and evaluating mitigation options. Most national greenhouse gas inventories rely on empirical Tier-1 or Tier-2 emission-factor approaches and therefore do not fully capture the effects of climate variability, soil properties, or management practices. Here, we present a transferable, process-based modelling framework based on the biogeochemical model LandscapeDNDC for determining direct and indirect N2O emissions from major crops cultivated on mineral soils at the national scale. We apply the method to Germany making use of high-resolution input data provided by the national reporting agencies, estimating N2O emissions of 35 (29–44) kt N yr-1(2017–2022 average). This is 28 % higher than the national inventory report (submission 2025), but well within the uncertainty range. In contrast to conventional inventory methods, the framework explicitly accounts for interannual climate variability and can be spatially disaggregated at high resolution, taking into account local variations in soil type, weather and agricultural management practices. Because the model simulates coupled carbon and nitrogen cycling, it also quantifies multiple nitrogen loss pathways and potential changes in carbon stocks simultaneously, providing a consistent basis for evaluating mitigation strategies and their potential trade-offs. Our results demonstrate that process-based modelling can substantially improve the spatial and temporal resolution of agricultural N₂O emissions and provide a platform for developing next-generation national greenhouse gas inventories. While further work is required before the framework fully satisfies all IPCC Tier-3 requirements, it offers a pathway towards a more mechanistic and policy-relevant assessment of agricultural greenhouse gas emissions.
- Preprint
(3446 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-4010', Anonymous Referee #1, 21 Aug 2026
-
RC2: 'Comment on egusphere-2026-4010', Anonymous Referee #2, 25 Aug 2026
Review of: "A process-based framework for national-scale estimation of agricultural soil N2O emissions under variable climate and management"
General comments:
The manuscript presents a process-based modeling framework for producing national and sub-national inventories of direct and indirect N2O emissions from agricultural soils (Tier-3), with a demonstration for Germany over 2017-2022, which is compared to the national inventory method (Tier-2). The study is well conducted, involves impressive work, and the manuscript is generally well written. I truly appreciate the efforts to assess the uncertainty of the model estimates as well as the careful comparison with the inventory data. However, I think some additional information and clarifications are important to improve the manuscript. My main concerns are the missing model evaluation, which leaves a big question mark on how much we can trust the estimates from this modeling framework, and secondly, the last calibration step involving the linear mixed effects model for model structural uncertainty needs clarification and more info in the Results and Discussion.
Major comments:- Model structural uncertainty calibration step
L125-132 This paragraph needs a better explanation. I am confused because it belongs to the calibration, but by reading these lines it seems this step had an effect only on calculated uncertainty. However, when reading L449-457, if I am correct, I understood that this step transformed all the predicted emission rates. If this is the case, there is a need for: i) a clearer explanation in L125-132 and L449-457; ii) reporting in the Results how this step changed the estimated emissions, by providing at least the average N2O emissions as predicted by LDNDC before this transformation; iii) in the Discussion, shortly discuss the relevance of this step and mention that this step makes this framework less transferable to regions with scarce N2O measurements, as they are needed to constrain the linear mixed effects model.
- Missing model evaluation
All 12 sites / 58 treatment-years / 8799 measurements are used in the calibration (L100-108). There is no hold-out set, no cross-validation, and no evaluation statistics of any kind. The manuscript therefore contains no evidence that the model was able to reproduce observed N2O fluxes. I strongly recommend adding such performance assessment, preferably holding out sites/years (no random split on all measurements to avoid time autocorrelation issues), and I would suggest showing the performance on both single measurements and cumulative seasonal/year emissions. Ideally, I believe it would be very interesting to also show the performance of the Tier 2 method, since you are comparing the estimates of both methods. I would also suggest showing the model performance for crop yields in the Appendix.
- Reduced uncertainty claim
I recommend better contextualizing the lower uncertainty achieved with the Tier-3 framework (L355-359 and L406). Although I truly praise the evident effort of including the uncertainty of the input data and of the model parameterization and structure, many uncertainties are not accounted for in the input data, and it’s hard to fully account for the inherent uncertainty of the model, as well as model bias. In addition, several educated guesses and filters have been used, which have likely reduced the estimated uncertainty. Therefore, I suggest framing these statements better by mentioning the complexity of including all the sources of uncertainty. Furthermore, if feasible, it would be interesting to know how much each uncertainty source contributed to the total uncertainty, by decomposing into e.g., inputs, calibration…
- Transferability of the modeling framework
The educated guesses and calibration steps, both requiring measurement data, reduce the transferability of the presented modeling framework, especially in data-poor regions/countries. Therefore, I recommend discussing this aspect in the Discussion and lowering the claim of being ‘easily transferable’.
Specific comments
L18. Actually, it is 28% lower than the recalculated NIR, not the published one. Please specify.
L48. Also need to mention weather variability alongside landscape variability.
L79. "In general, there is one harvest per year" is not really representative since you include grasslands. Related to this, throughout the paper "NIR (sub 2025)" is used as a label for values computed by re-applying NIR methods. This can be confused with the officially reported inventory. Please relabel, e.g. "NIR method (recomputed)".
L97-98. “When running…weights” doesn’t read well; please rephrase.
L108. Not clear what you refer to with “model species parameters” (could be crops, nitrogen forms…)
L118. Need to explain here or refer to the Appendix because I would not know how you judged a “physically reasonable relationship”
L136-139. It would be relevant to get an idea of the distribution of these sampled points. At the moment we don’t know how many points the small districts are getting, which might have quite an impact on the estimated emissions due to lower sampling. I would suggest adding a Figure or Table in the Appendix to show the distribution and refer to this in the main manuscript. In particular, I think it would be valuable to mention this in L294.
L143-147 and L189-194. Why did you subtract only the reduced form of deposited N?
L224-227. A 382 vs 250 difference between the two estimates is a big discrepancy and needs some discussion/explanation.
L327. “uses a model intercept” needs to be explained better, and please add a reference.
L331-336. The paper is already very long, so this is just a suggestion: it would be very interesting to see if the simulated N surplus vs N2O emissions relationship is not linear; you could add a Fig. to the Appendix and refer to it here to make your statement stronger.
L340-341. Please argue the reasons for the site emissions being lower than the simulated German average.
L344-346. You start with “the greater interannual variation” and then you say the opposite “one surprise…does not show larger interannual variation”. I think in the second sentence you wanted to refer only to 2018; please clarify and make it more consistent.
L369. “will likely never be covered” is a strong statement; please argue why or change.
L396. Please mention the higher uncertainty when using global datasets.
L406-409. Your work does not demonstrate the model capability of simulating contrasting management practices, so I would not go in this direction. I suggest deleting this sentence.
Figure 2. I suggest showing only the indirect emissions in the lower panel since they are small and when summed with direct emissions it’s hard to judge their magnitude and uncertainty.
Figure 4a. I suggest coloring the bars with two different colors depending on the method (Tir-3 vs NIR; add color legend) so that you can keep the pattern for both bars/methods. As the climatic zone is specified below, there is no need for different coloring for that. At the moment is not very intuitive.
Table D1. Add a column giving the measurement method (static/automated chamber, or eddy covariance).
Technical corrections/suggestions
L30. “emission fluxes” is redundant
L59. “allow” to “allows”
L78. “land area” to “total land area”
L207. “and receive” to “that received”
L344. “to a higher sensitivity” to “to the sensitivity” because EFs are not sensitive at all.
Other
There are two sections numbered 2.4 and no 2.3. In the Results, there is a 3.1 and a 3.3 but no 3.2.
Section 4 is very long, so I believe that breaking it into subsections would help the reader
In a new Appendix Fig or in a current main figure with the map of Germany, add the pins for the sites used for model calibration
Citation: https://doi.org/10.5194/egusphere-2026-4010-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 172 | 54 | 15 | 241 | 21 | 21 |
- HTML: 172
- PDF: 54
- XML: 15
- Total: 241
- BibTeX: 21
- EndNote: 21
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of: "A process-based framework for national-scale estimation of agricultural soil N2O emissions under variable climate and management”
General comments
The manuscript presents a Tier-3 N2O estimation framework from agricultural soils that can be applied using a multitude of global- or national-level grid-scale datasets. The study is timely and relevant to climate science and mitigation strategies, particularly important as a greater number of countries aim to reduce their N2O emissions from agriculture. The study shows that Tier-3 emissions are 28% higher than national emission inventories, but are within the uncertainty bounds, with the added advantage of high temporal resolution. However, this comes at a cost of computation and expert knowledge, which is not always readily available globally. Overall, the manuscript was very enjoyable to read, and the authors have produced a high-quality manuscript.
Specific comments