Review of "ORCHIDEE-PEAT Model and Peatland Drainage Experiments"

Summary:

This study calibrates the ORCHIDEE-PEAT model at ten pristine peatland sites to simulate CO₂ and CH₄ emissions over multiple years under these pristine conditions. Based on the calibrated parameters, the authors conduct hypothetical drainage experiments to investigate how greenhouse gas fluxes change when the water table is lowered. The research aims to address the transitional phase from pristine to drained peatland conditions—a critical phase that is often missing in observational datasets.

The idea is great to fill this knowledge gap with a model-based approach, however, I have serious concerns about how the modeling approach is currently implemented. The key issue is that the calibration is based only on pristine conditions, but the model is used to predict emissions in drained conditions, where apart from water level which is lowered by drainage other environmental controls are fundamentally different. The model does not account for critical processes such as vegetation shifts, changes in soil organic matter, and agricultural practices like biomass removal or fertilization following drainage. These omissions raise fundamental doubts about the reliability of the simulation results and the conclusions drawn by the others. 

One highly questionable conclusion of this study is the claim that it provides "a more nuanced view than the current paradigm that drainage always warms the climate." Measurement data indicate that peatland drainage results in long-term climate warming. It is well established that the time horizon is crucial, as the radiative forcing of long-lived greenhouse gases (e.g., CO₂) is driven by cumulative emissions, whereas the radiative forcing of short-lived climate forcers (e.g., CH₄) depends on contemporary emission rates. Meta-analyses show that, after just one or a few decades, the net impact is unequivocally climate warming. (https://doi.org/10.1038/s41558-019-0615-5, https://doi.org/10.1038/s41467-020-15499-z).    

Below, I outline my main concerns in more detail.

Major Comments:

1) Unjustified extrapolation beyond the calibration range

The model is calibrated using data from pristine peatlands, but then applied to drained conditions, environmental conditions are entirely different. Using the same parameter set for both conditions is a major extrapolation and likely introduces massive errors.

A better approach would involve calibrating the model using both pristine and drained sites. Multi-site calibration covering the full range of conditions would improve the robustness of the model and reduce uncertainty in the predictions. Without such an approach, the results cannot be intepreted in the context of real world situations. 

2) Lack of realistic vegetation changes

In reality, peatland drainage always leads to significant changes in vegetation. Many drained peatlands are converted into cropland or grassland, fundamentally altering the plant community, biomass inputs, and ecosystem functioning. However, the model assumes that only water levels and soil moisture change, while vegetation remains the same, i.e. pristine vegetation is assumed to persist under drained conditions, which is cleary impossible for many peat-specific species.

The authors acknowledge this as a limitation, which is good, but I believe it is too severe to allow meaningful interpretation of the results. The fact that model outputs match the range of measured data from drained sites does not justify the approach if the model assumes completely different conditions. The fact that net ecosystem exchange (NEE) values fall within observed ranges does not mean the model is capturing the right processes, it could simply be matching for the wrong reasons.

Moreover, the model does not account for biomass removal, which is critical when peatlands are converted to agriculture, neither for the strong impact of nutrients on peat decomposition. Without incorporating realistic vegetation and land-use related changes, the results cannot be meaningfully compared to real-world drained peatlands.

3) Incomplete representation of soil organic matter (SOM) changes

The study attributes the simulated decline in CO₂ emissions over time to the depletion of labile carbon pools. This is not in line with measured datasets and seems to be a model artifact. The model does not seem to account for the fact that SOM composition changes with ongoing decomposition. Decomposed peat is often more vulnerable to further breakdown than less degraded peat. Observational studies indicate that drainage increases the portion of highly decomposed organic matter, which can sustain CO₂ emissions over longer periods (https://doi.org/10.1016/j.geoderma.2019.113911). If the model does not represent this change, then the decline in emissions might not be a realistic outcome but rather an artifact of how SOM dynamics are treated.

4) Weak validation approach

The validation method used in the study is not robust enough to assess the model’s reliability. The so-called “80/20” split seems to rely on arbitrary data partitioning, possibly in seven-day blocks. Given the strong temporal autocorrelation in peatland flux datasets (also for seven-days), such an approach can lead to overestimated model performance (https://doi.org/10.1111/ecog.02881). More rigorous validation strategies, i.e. a more systematic approach to separating training and validation datasets, should be used. Independent time periods or entirely different sites should be used for validation.

5) Unclear rationale for using an ML-based water level model

Previous studies using ORCHIDEE-PEAT have simulated peatland water levels using process-based approaches. However, in this study, a machine learning (ML)-based model is used instead. The reason for this shift is not well explained, nor is the ML model sufficiently validated. What advantages does this approach offer over a physically based model? Without a strong rationale and proper validation, it is difficult to assess whether the ML approach improves or weakens the reliability of the results. More transparency on this choice would help clarify its impact on the findings.

6) Insufficient explanation of the drainage module

The study mentions the implementation of a drainage module, but it is unclear what exactly this entails. One key issue is whether it realistically represents soil moisture dynamics in the unsaturated zone. Capillary forces play a crucial role in maintaining moisture levels above the water table, and ignoring them can lead to incorrect soil moisture predictions. The paper does not make it clear whether this critical process is accounted for. A more detailed description of how the drainage module works with details on the underlying physics of the soil hydraulic approach would improve clarity.

Minor Comments

Line 40: Given the limitations of the study, the claim made here seems overly optimistic. Since the authors themselves acknowledge several key limitations, this statement is sufficiently supported by the study results. 

Lines 66–70: The text suggests that only “some” studies report long-term CO₂ emissions after drainage, but in fact, the vast majority of research supports this finding. Evidence from drained peatlands in the UK, Netherlands and Germany, where meters of peat have been lost (compaction alone cannot explain this) over time evidenced by timber posts (e.g. https://www.greatfen.org.uk/about-great-fen/heritage/holme-fen-posts, many more exist), contradicts any suggestion that emissions decline significantly in the long run if there is still peat available to be oxidized.

Line 71: The terminology and sign convention regarding water table level and water table depth is inconsistent throughout the manuscript. This should be standardized to avoid confusion about what high, low, deep, shallow, etc. means.

Final comments:

I think the focus on the transitional phase between pristine and drained peatland conditions is valuable, as this is an important and underrepresented topic in peatland research. However, the modeling approach has fundamental weaknesses that limit the reliability of the findings.

I hope these comments are taken in the constructive spirit in which they are intended. Improving these aspects in a fundamentally revised paper will help strengthen the study and ensure that its conclusions are robust.

Reviewer 1 has provided a thorough and insightful evaluation of the manuscript, highlighting several key issues that warrant attention. I won’t repeat all their points here, but I would like to express my full agreement with their assessment.

That said, I also share Reviewer 1's view that the paper addresses an important topic—modeling peatland drainage and its effects on greenhouse gas emissions. This is a valuable area of research that deserves continued development. The modeling approach appears to perform quite well under pristine conditions, especially where the water table depth (WTD) remains above -20 cm in the calibration datasets. In that respect, the paper serves as a strong example of model development. It may also serve to inform of what may happen to pristine systems as these are affected by warmer and drier climates.

However, the system being modeled—drained peat soils—is highly complex. These soils behave quite differently from mineral soils or even undisturbed peatlands. Lessons from better-studied systems may not always transfer well, especially for peatlands that have been drained for long periods.

My primary concern lies in some of the claims made based on this model. These are bold conclusions, and in such cases, it's essential to ensure that they are supported by robust evidence. At present, I’m not fully convinced that the data and analysis sufficiently back these claims.

There are many innovative aspects in the manuscript, but the results hinge critically on the assumption of a slow SOC pool in deeper peat layers. As Reviewer 1 has pointed out (and I had also intended to reference), this assumption may not reflect reality. In fact, evidence suggests that in agricultural peatlands, decomposition rates often increase over time as the peat becomes more degraded. This point should be reconsidered in future model iterations.

I'd also like to raise a few additional considerations:

1. Drainage depths used in the model runs: The levels tested (-5, -10, -20, and -50 cm) seem quite shallow. Only the -50 cm scenario might resemble an initial drainage condition. Including deeper, more representative drainage levels could provide more realistic insights.

2. Dynamics of managed drainage systems: In real-world settings, drainage in peat soils is often actively maintained over time—ditches are deepened, compacted soils are re-drained, etc. This means that drainage may reach progressively deeper peat layers, encountering different SOM characteristics than those present at the start. The model could reflect this temporal change more explicitly.

3. Use of GWP for interpreting results: While GWP is commonly used in policy contexts, it may not always capture the complexity of peatland carbon dynamics—particularly the long-term CO₂ storage versus CH₄ emissions trade-off. Radiative forcing models might offer a more nuanced picture and could be used to contextualize the GWP results more effectively.

A few minor suggestions:

Please clarify whether the point data in the figures represent actual site measurements or model outputs.

In line 495, it seems there may be a typo in the equation: should it be intNEE rather than intCH₄?

With a more realistic representation of drained peat soils, this modeling framework has strong potential to inform both research and policy. I hope these suggestions are helpful and supportive of the manuscript’s continued development.