This study provides a unique millennium-scale perspective on land-use change (LUC) emissions in China, addressing critical gaps in reconstructing historical LUC data and updating contemporary emissions modeling. While the data and modeling are not perfect at this point, the study has made great improvements to LUC data since the 1000s, and updated carbon densities for current biomass and soil. The manuscript is well-structured and easy to follow, but its contributions and methodological choices require further clarification to strengthen its impact. I would recommend publication after revisions.

Major Concerns:

The study’s novelty should be explicitly contextualized. Why is a millennium-scale analysis of LUC emissions critical, given the inherent uncertainties in pre-industrial data? How does this long-term perspective enhance our understanding of anthropogenic impacts on carbon cycling, even when CO2 levels were relatively stable before industrialization? China’s uniquely long historical record enables this work, but how might its findings inform global LUC emission estimates, particularly for regions with limited historical documentation?

Regarding LUC data: It is challenging, if not impossible, to validate the LUC over the past millennium. The “reliability assessment” of historical LUC data needs elaboration. How does this assessment validate the reconstructed data, given the absence of direct validation methods for pre-industrial periods? Clarify whether this approach evaluates internal consistency, cross-references with alternative proxies (e.g., tax records), or quantifies uncertainty ranges. Please explicitly state what distinguishes the LUC dataset in this study from prior publications by He et al. Is the novelty in data synthesis, spatial resolution, or integration of new historical sources (e.g., tax records)?

Regarding carbon density assumptions: The assumption of static carbon densities over millennia is problematic. While the authors update current biomass and soil densities, pre-industrial carbon stocks likely shifted due to CO2 changes, climatic variability, ecological succession, and human management. Discuss how these dynamics might bias emission estimates and propose strategies to address this in future work (e.g., coupling with DGVM outputs). The carbon density updates in the current work only scratched the surface of the issue, by improving the densities of “current” times. In GCB2024, there are four book-keeping models used, why do you choose H&N or H&C model (I assumed, you did not specify)? Is it because of spatial resolution or any particular features that match well with your current data, like using LUC “state” instead of LUC “transition”? The other three seem to incorporate dynamic carbon densities to some extent (for instance including DGVM biomass data), but also with higher spatial resolution that may not match the provincial level in this study. I would suggest clarifying the rationale in the Methods, AND further discussing the uncertainties in the Discussions. This is not to deemphasize this work, but to urge future improvements.

About uncertainty quantification: The current “uncertainty” section (4.3) primarily discusses limitations rather than quantifying uncertainties. Incorporate a robust quantitative analysis (e.g., Monte Carlo simulations) to assess how data gaps (e.g., historical LUC, carbon density variability) propagate into emission uncertainties. This will enhance the study’s rigor and reproducibility. The 4.3 section is not technically an “uncertainty analysis”, it is simply discussions of limitations and possible future work.

Minor points:

“China”: this needs to be better defined in this study! You used the current mainland China as the country boundary, and merged the 30+ provinces into 25 regions. I understand your reasoning for compromising here, but you must make this crystal clear in the Abstract and Methods. In Fig. 1, you may also cite specific studies for each map for different dynasties.

L171: regarding the bookkeeping model, did you use the Houghton model, or simply used their structure and data? This can be made more explicit.

L200: “local expert and knowledge”, delete “and”?

L206: using tax records is a great idea, but how does this help this particular study? Any quantitative evidence?

L224: this is out of context, what exactly is “inverted S-shaped” relationship?

L243: cite the data used.

L270: Fig 3. The whole study is at the provincial level, why do you use gridded data here in the map? What data are they? What criteria did you use to separate west vs. east of China, or to draw the “forest-grassland boundary”? Over 1000 years, did this boundary move at all?

L290: the whole argument about shifting ag. in China is not strongly supported. This happens in Africa and S. America, but it is not as common in China. What does recent remote sensing suggest? It would be more convincing to show some direct evidence than simply claim “…has been recorded extensively in Chinese historical documents.”

Fig. 4-5, did you compare the LUC data with other sources, like LUH2, to examine the differences and causes?

Fig. 5: please clarify the meaning of secondary axis. In (a), does the y-axis suggest “changes” or absolute area? Same for (b), absolute or relative area? For (c) and (d), what does the pie suggest, 1000-yr cumulative or annual?? Please be more specific.

Fig. 6: Does the negative biomass value show carbon sink? Specify in the caption.

L435: Table3, this table is a summary not “comparison. These estimates cover different time period, so the emissions would be different. No surprise here. Could you compare them across the same or similar time, and include results from this study?

L476: Is this required? It seems odd with a data availability statement in the middle.

Appendix A and B: is the information in these tables used in this study? Or do they simply support previous work on LUC data.

Publisher’s note: the content of this comment was removed on 20 May 2025 since the comment was posted by mistake.

This manuscript presents a millennial-scale reconstruction of carbon emissions from land-use change in China using a bookkeeping model approach. While the study addresses an important research gap and provides valuable historical context for understanding China's carbon budget, there are several major concerns that must be addressed before this work is suitable for publication.

1. The conversion rules in Figure 3 appear somewhat arbitrary. I recommend testing the uncertainty in your transition matrix calculations. While your rule-based priority system is clear, how would results differ if you used an area-weighted approach instead? For example, allocating transitions proportionally based on the relative magnitude of area changes between different biomes rather than using predetermined priorities. This uncertainty analysis would be valuable given the millennium-long timeframe of your study, where even small methodological differences could compound into significant variations in results.

2. The authors state that “this study updated and improved the land-use change data, carbon density data, and disturbance response curves,” but upon careful reading, it appears they did not actually update or improve the disturbance response curves themselves. Rather, they simply adopted the data from Houghton and Castanho (2023) without modification. To avoid misleading readers, I suggest the authors clarify that they utilized the most recently published parameters from the literature rather than implying they developed improvements to the response curve themselves.

3. I also noticed that the bookkeeping model used in this study does not account for wood harvest pools, which is understandable given that it would require reconstructing additional historical wood harvest data. However, this limitation should be explicitly stated in the methodology section. The authors should clarify this omission and briefly discuss its potential implications for carbon flux estimates, especially since wood harvest can be a significant component of land-use change emissions in forested regions of China. 

4. The explanation of differences between NGHGI and bookkeeping estimates should focus on carbon accounting boundaries rather than restoration projects (Gidden et al., 2023, Nature; He et al., 2024, Nature Communications). For DGVMs vs. bookkeeping models, note that DGVMs include the loss of additional sink capacity, leading to higher emission estimates, alongside differences in LUC forcing data (Gasser et al., 2020, Biogeosciences). I suggest the authors provide a more systematic discussion to avoid misleading readers about these differences.

Manuscript Title: Annual carbon emissions from land-use change in China from 1000 to 2019

Recommendation: Major Revision

General Comments

This manuscript presents an ambitious reconstruction of carbon emissions from land-use change (LUC) in China over the past millennium. Using a provincial-scale bookkeeping model and extensive historical records, the authors estimate annual LUC emissions from 1000 to 2019, supported by updated carbon density datasets. The work contributes a long-term dataset of carbon fluxes that could support both paleoclimate-carbon research and national greenhouse gas (GHG) accounting.

However, the manuscript falls short in clearly articulating its scientific motivation, ensuring methodological transparency, and validating the results. Of particular concern is the assumption that vegetation and soil carbon densities remain constant over 1000 years, which critically weakens the interpretability of the results. In addition, the complete absence of quantitative uncertainty analysis and comparison with existing datasets limits the credibility and broader applicability of the findings.

I recommend major revision to address the following concerns.

Major Comments

1. The scientific rationale, challenges, and innovation of a millennial-scale reconstruction are insufficiently articulated

While reconstructing LUC-related carbon emissions since AD 1000 is conceptually valuable, the manuscript does not sufficiently explain:

• Why this timescale is necessary for understanding anthropogenic impacts on the carbon cycle;
• What methodological or conceptual challenges exist in performing such long-term reconstructions;
• How this study specifically overcomes those challenges or improves upon prior work.

The novelty of the study must be made more explicit. For example:

• How does this reconstruction differ from studies that begin in 1700 or 1850?
• What new historical sources, spatial refinements, or analytical methods are introduced?

Recommendation: Include a comparative table summarizing key differences between this study and prior LUC carbon emission reconstructions (e.g., in time span, resolution, input data, model approach, and validation).

1. The assumption of static carbon densities undermines the long-term credibility of the reconstruction

A central concern lies in the assumption that vegetation and soil carbon densities remain constant over the entire 1000-year period. While this may be a necessary simplification given limited historical data, it significantly weakens the scientific credibility of the estimated carbon fluxes—especially for earlier centuries.

Carbon densities are not time-invariant: they are influenced by changes in climate, atmospheric CO₂, ecosystem succession, species composition, and land-use intensity. Assuming present-day carbon densities for all historical periods’ risks introducing systemic bias in the emission estimates, particularly during major climatic or socio-ecological transitions (e.g., the Little Ice Age, or the Qing Dynasty agricultural expansion).

This assumption is particularly problematic because the technical challenge—and scientific value—of millennial-scale carbon accounting lies precisely in addressing such temporal variability. If a key driver like carbon density is held static, the study risks becoming an arithmetic exercise rather than a meaningful reconstruction, and its findings may not substantially differ from earlier studies based on heuristic extrapolation.

Recommendation:

• Clearly state which carbon density datasets are used, and how they are applied across the time domain;
• Acknowledge the limitations of assuming static carbon densities, and discuss the potential magnitude and direction of bias this may introduce;
• Propose a pathway for future work, such as incorporating carbon density outputs from process-based vegetation models (e.g., DGVMs) or paleoecologically reconstructions;
• Emphasize that confronting this assumption is essential for enhancing the interpretive value and novelty of the study.

1. Modern-era results lack validation and comparison with existing datasets

The study spans from 1000 to 2019, but observational and model-constrained datasets are available primarily for the post-1950 period. Yet the manuscript does not compare its estimates to:

• National or global LUC carbon emission inventories (e.g., FAO, Houghton, LUH2);
• Remote sensing-based datasets of forest loss or biomass change;
• Process-based models such as DGVMs or spatially explicit bookkeeping models (e.g., BLUE).

These comparisons are essential for establishing the reliability of the methodology and providing a reference point for earlier trends.

Recommendation: Include a table comparing national and/or provincial LUC emissions from this study with at least 3–4 widely used datasets over overlapping time periods, accompanied by discussion on differences and their likely causes.

1. Absence of quantitative uncertainty analysis limits credibility

Section 4.3 is labeled “Uncertainty Analysis” but provides only qualitative reflections on limitations. This is insufficient given the range of assumptions, spatial heterogeneity, and sparse data for earlier centuries.

Recommendation:

• Include a quantitative uncertainty analysis (e.g., via Monte Carlo simulations or scenario analysis);
• Report confidence intervals or uncertainty bounds for cumulative and decadal emissions;
• Indicate how uncertainty varies across time, especially between well-documented (post-1950) and poorly constrained (pre-1700) periods.

 Minor Comments

1. Clarify the bookkeeping framework
   
   Indicate whether this is a “statistical bookkeeping model” or incorporates spatially explicit components to distinguish it from models such as BLUE or OSCAR.
2. Add a conceptual model diagram
   
   A schematic showing the flow from land-use data → transition → response curve → carbon flux would clarify the modeling approach.
3. Improve Table 2
   
   Include the number of observations or sample density for each province to help readers assess data quality.
4. Clarify carbon density preprocessing
   
   Indicate whether carbon density values were standardized (e.g., by reference depth) and whether outliers were removed.
5. Enhance regional time-series presentation (e.g., Fig. 7)
   
   Add temporal trends for individual regions, not just cumulative bar plots.
6. Clarify the role of Appendices A and B
   
   State whether the sources listed in the appendices were used directly in this study or referenced for historical context only.
7. Ensure consistent terminology
   
   Maintain consistent use of terms like “carbon sink” vs. “carbon sequestration.”
8. Improve figure quality
   
   Enhance the resolution of Figures 5–8 and define all abbreviations in figure captions (e.g., “Yue-Qiong”).
9. Refine the title for clarity
   
   Consider including terms such as “provincial reconstruction” or “bookkeeping-based estimate” to better reflect the methodological approach.