Singh et al. “Modelling framework for asynchronous land-atmosphere coupling using NASA GISS ModelE and LPJ-LMfire: Design, Application and Evaluation for the 2.5ka period”

GENERAL

This study simulates the climate of the 2.5 ka period using the NASA GISS ModelE, which is asynchronously coupled to the dynamic vegetation model LPJ-LMfire. The authors conducted several sensitivity experiments to assess the impacts of initial land cover and bias correction on paleoclimate and ancient vegetation. Additionally, they used multiple proxy datasets to validate the simulation results. While this study represents a considerable effort in developing a modeling framework and conducting numerical simulations, the robustness of the main results is questionable, and the study's novelty remains unclear.

First, the study lacks sufficient novelty. The primary focus is on asynchronous land-atmosphere coupling between a climate model and a vegetation model. However, it is unclear why this coupling approach warrants extensive application, as it does not appear to enhance the simulated paleoclimate. As shown in Fig. 12, temperature anomalies exhibited the lowest biases in the 2.5k_PI_ctrl simulation, which did not include asynchronous land-atmosphere coupling. This finding suggests that vegetation coupling has a minimal impact on improving model performance, thereby weakening the study's novelty.

Second, the key sources of discrepancies remain unexplored. While the authors compared results from multiple experiments, they primarily noted that bias correction led to a warmer and more stabilized climate, without further investigating the underlying causes of climate variations across different simulations. In particular, the influence of initial vegetation cover, climate variability, and climate-vegetation interactions on the simulated paleoclimate should be explicitly quantified and explained. Furthermore, the δ18O validations revealed similar biases across different experimental configurations (Fig. 15), raising the question: what is the added value of employing different model configurations?

SPECIFIC

Line 40: “The coupled model system is sensitive to the representation of shrubs.” What is the underlying reason for this particular sensitivity? Is this a characteristic of this specific model, or is it a general feature of all coupled models?

Lines 176-177: How well does the LPJ model perform in simulating present-day vegetation? It would be more appropriate to validate the LPJ model using observed meteorological data and vegetation parameters rather than relying solely on PI simulations.

Line 207: What observational datasets could be used to validate the derived ‘wet days’? Additionally, models tend to overestimate the frequency of small rainfall events, which suggests that the nonlinear relationship derived from observations may not be directly applicable to model outputs.

Line 220: “Adding interannual variability”—why is this an important consideration? How do simulations differ with and without interannual variability?

Line 235: What is the connection between plant functional types (PFTs) and fire frequency? Please clarify the underlying mechanism.

Line 243: “A predefined threshold”—Is this threshold applied uniformly at the global scale, or do different grid cells use varied thresholds? How was this threshold determined?

Line 272: “climate vegetation models”—this phrase is missing an “and” (should be “climate and vegetation models”).

Line 285: “Linearly interpolating”—Are there any proxy datasets available to constrain vegetation cover for the 2.5 ka period? How would different initial conditions affect the final simulated vegetation cover? Additionally, how can the derived paleo-vegetation be evaluated?

Table 2: The reasoning behind the different simulation lengths across various experiments is unclear. What criteria were used to determine the duration of each experiment?

Figure 3: Is it reasonable to use present-day satellite-derived land cover for PI simulations? Please provide justification.

Lines 389-391: These sentences belong in the figure captions rather than the main text.

Figure 10: The differences between LPJ and GISS, as well as between PI and GS, are relatively minor. The most significant differences appear between x and BC. This conclusion should be explicitly stated and explained in the main text.

Figure 11: In some regions, differences among simulations are substantial, while in others, they are minimal. What accounts for these regional variations? What role does vegetation play in shaping these differences?

Section 5: This section should be positioned before the analysis for better logical flow.

Figure 12: Proxy data appear to be more consistent with PI simulations than with the 2.5 ka simulation. What is the purpose of bias correction or climate-vegetation coupling if it does not improve the agreement with proxy data?

Line 640: How is isotopic composition represented in the model simulations? Please provide details on its setup.

This manuscript introduces an asynchronous coupling framework between the NASA GISS ModelE atmosphere/climate model and the dynamic global vegetation model LPJ-LMfire. The framework is applied to mid- to late-Holocene (2.5 ka) simulations, exploring how dynamic vegetation and fire affect vegetation cover, albedo, and regional climate. The idea is timely and valuable: asynchronous coupling between atmosphere models and DGVMs can help advance palaeoclimate simulations and improve the realism of land–atmosphere interactions. However, it lacks sufficient novelty. In its current form, it lacks sufficient methodological transparency, reproducibility, and robustness in validation to meet GMD standards.

Major comments:

1. The coupling strategy must be documented in a way that allows others to reproduce it. The current description is too brief. Please provide a step-by-step description of the asynchronous loop (order of model runs, frequency of exchanges, number of iterations). A complete list of exchanged variables, with units, spatial grids, interpolation method, temporal aggregation, and post-processing. Table 1 can be expanded to contain these details, or a detailed flow chart can be added. Any bias correction, spin-up, or equilibration applied between iterations. How did you validate the model results fit with reality? All of these need to be shown and discussed.
2. The manuscript does not clearly specify how many experiments/ensembles were conducted, nor their lengths. Please add a concise experiment table with experiment names, forcings applied, spin-up duration, ensemble size, and control/baseline. Clarify whether ensemble spread reflects internal variability or parameter uncertainty.
3. State explicitly which CO2/greenhouse gas values and orbital forcings were used for each experiment, with citations to PMIP4 or other sources. A dedicated table listing GHG concentrations would be helpful.
4. The choice of monthly coupling is not justified. Either show sensitivity to coupling interval (e.g., monthly vs seasonal/annual) or provide a discussion, with references, of the limitations and expected impacts of this choice.
5. The manuscript compares simulations with reconstructions but lacks detail on datasets and metrics. Please list all proxy datasets used, with version numbers and coverage. Describe the model-data comparison method (regridding, seasonality, temporal windows). Provide quantitative metrics (bias, RMSE, correlation, significance tests) and report proxy uncertainties.
6. Discuss uncertainties more systematically: (i) forcings, (ii) internal variability, and (iii) mapping of LPJ PFTs to ModelE classes. If multiple mapping schemes were tested, please include the results or add them to the Supplement.
7. Clarify how fire is represented in LPJ-LMfire for these experiments: which ignition sources are included, how burned area is calibrated, and whether fire emissions feed back to ModelE. If emissions are excluded, state this explicitly.
8. Improve figure captions so each is interpretable without referring to the main text. All figures must include units, baseline (difference relative to what), averaging period, and statistical significance.

Minor comments:

1. Be consistent with "2.5 ka" wording – clarify whether you mean 2.5 ka BP (2500 years before present), 2.5kyr, or "mid-Holocene" and give exact calendar years used.
2. The manuscript lists a few CO2 values (e.g., 271.4 ppm, ~279 ppm, 284 ppm). Make sure each value is linked to the correct experiment and source citation. Possibly a single table listing GHGs for each experiment would be useful.
3. You mention testing several LPJ-GISS PFT mapping schemes. Please show results (or at least diagnostics) illustrating how sensitive your key climate metrics are to the mapping choice.
4. A few citations seem incomplete or inconsistently formatted. Please check the references for completeness and adhere to the journal style.
5. In the discussion, the text implies strong climate impacts from the coupling that are not fully supported by the presented evidence. Tone down causal language where results are suggestive rather than robust and clearly label speculative statements.