Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

An integrated simulation framework combining an improved gray prediction model with the CLUMondo spatial model was developed in this paper. The simulation framework and policy insights proposed herein offer valuable scientific reference and decision-making guidance for sustainable land use and refined spatial governance in Ningxia. However, it requires further improvement to meet the requirements for publication. Specific revisions are suggested as follows:

1. What are the differences between the cropland change simulation of arid region and that of other regions?
2. Table 5: Why were only the conversions among these five types of land considered? What is the basis for type classification?
3. Section 3: The simulation process needs to take into account the impact of policies. For instance, high-standard farmland is usually not allowed to be converted into other land use types.
4. Section 4: It is suggested to conduct a further in-depth comparison with other simulation methods.
5. Section 2.5.4: What is the basis for choosing Regression Parameters? Have the characteristics of arid regions been taken into consideration?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Some comments here:
(1) While the manuscript proposes a coupled framework integrating an improved GM model with the CLUMondo model, the methodological novelty remains insufficiently differentiated from existing LUCC simulation studies that also combine demand forecasting and spatial allocation. The improvements to the GM model (metabolism mechanism and residual correction) are technically sound but appear incremental, and their broader applicability beyond the Ningxia case is not fully demonstrated. Therefore, the authors should more explicitly benchmark the improved GM model against alternative demand-forecasting approaches (e.g., Markov chains, SD-based models, or machine-learning regressors) using quantitative error metrics. Additionally, a clearer discussion is needed on how this coupled framework can be transferred to other arid or semi-arid regions with different policy regimes, data availability, and land system dynamics, thereby strengthening the paper’s contribution to land system science rather than a single-case application.

(2) The manuscript emphasizes the incorporation of policy factors (e.g., ecological migration, cropland protection, ecological compensation) as a key innovation. However, the operationalization of these policy variables in the CLUMondo model remains partly qualitative, and their empirical validation is limited. I would like to suggest the authors provide a clearer explanation of how each policy factor is quantitatively translated into model parameters (e.g., regression coefficients, resistance values, restricted areas). Where possible, policy implementation data (such as spatial extents, timelines, or funding intensity) should be used to justify parameter settings. Sensitivity or scenario perturbation analyses focusing specifically on policy-related parameters would also help demonstrate the robustness of the conclusions regarding policy effectiveness.

(3) Although three development scenarios (BAU, EP, and URE) are well designed and intuitively meaningful, the manuscript treats scenario outcomes largely as deterministic results. Uncertainty arising from parameter choices, data resolution, and future socioeconomic trajectories is not sufficiently addressed.

(4) The discussion section primarily reiterates results and focuses strongly on regional policy implications for Ningxia, while broader theoretical implications for land-use modeling and land system governance are underdeveloped. This limits the manuscript’s appeal to an international readership of Land. The authors should strengthen the discussion by more explicitly linking their findings to international literature on arid-region land systems, cropland protection, and spatial planning under ecological constraints. A comparative reflection on how Ningxia’s policy-driven land dynamics resemble or differ from other global dryland regions would improve the manuscript’s general relevance.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript proposes an integrated cropland-change simulation framework for an arid, ecologically fragile region (Ningxia, China) by coupling an improved grey prediction model (Improved GM(1,1)) for land-demand forecasting with the CLUMondo spatial allocation model. It analyses historical cropland change and simulates future land-use patterns under three scenarios. The topic is suitable for the journal, but several issues must be addressed:

1. The introduction is well written and establishes the general LUCC context, but it requires tightening and stronger scholarly grounding. The aim and scope need clearer operationalisation (there is no specific research question), and the quality of the literature review is inconsistent (the reference list includes items that appear unrelated to LUCC or cropland simulation).
2. Add one or two paragraphs summarising recent cropland or LUCC scenario simulations in arid China. From this, identify exactly what is missing and state two to four research questions.
3. In the methods section, the training window length, parameter estimation approach, evaluation metrics, and a benchmark comparison against baseline demand methods are missing. The paper claims improved precision but does not provide quantitative evidence of improvement.
4. The paper states that non-spatial statistical indicators were “spatialised through downscaling techniques” and that policy factors (subsidies, eco-migration, cropland protection, etc.) are included, but it lacks explanations of the exact downscaling method, the spatial units used, and how “policy rasters” were constructed.
5. There is inconsistency in the study period: the abstract and results refer to 2009–2024, while the conclusion repeatedly refers to 2009–2020.
6. There is inconsistency regarding optimal resolution: the text reports that 50 m has the highest AUC (Fig. 6 discussion), but elsewhere states that 100 m × 100 m is the “optimal” resolution.
7. There is inconsistency in accuracy metrics: the abstract and conclusion cite Kappa = 0.87 and OA = 91.72%, while Table 6 shows different OA values (90.34% and 93.34%) and Kappa around 0.85.
8. Table 3 appears incomplete and should be incorporated into the text rather than presented as a table.
9. Several figures appear dense with small labels and may be difficult to read in print or PDF. Ensure all maps consistently include a legible legend, scale bar, north arrow (where appropriate), a consistent colour palette across scenarios, and readable font sizes.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

General comment:

The manuscript addresses an important topic related to cropland change simulation in arid and ecologically fragile regions. The integration of an improved GM(1,1) model with the CLUMondo framework is relevant and potentially valuable. The manuscript is generally well structured, but several methodological descriptions lack clarity.

 

Specific comments:

 

1. Figures 4contain small text, unclear legends, or insufficient explanation of symbols.
2. The manuscript does not provide tables summarizing model parameters, transition matrices, or scenario assumptions, which limits transparency.

 

3. The improved GM(1,1) model is described conceptually, but the exact parameterization, validation metrics, and residual correction procedures are not fully documented. The metabolism mechanism requires clearer justification and reproducible steps.

 

4. The manuscript reports AUC, Kappa, and OA values but does not provide confidence intervals or cross-validation details. The logistic regression used for suitability analysis lacks information on multicollinearity testing, variable selection criteria, and model diagnostics.
5. The BAU, EP, and URE scenarios are described qualitatively, but the quantitative constraints, policy assumptions, and land-demand trajectories are not explicitly tabulated. This limits reproducibility and interpretability.
6. The introduction emphasizes policy-driven land-use transitions, yet the results section does not sufficiently quantify policy impacts or compare scenario outcomes using measurable indicators.
7. The manuscript relies on multiple administrative datasets, but the exact years, preprocessing steps, and potential biases (classification inconsistencies) are not fully addressed. Several sentences are overly long or contain grammatical errors. Some sections repeat information. The narrative could be more concise
8. The study does not discuss uncertainties related to socioeconomic projections, policy implementation variability, or the influence of external drivers such as climate change.
9. The manuscript mentions 20,000 iterations but does not explain how convergence was assessed or how transition elasticity parameters were calibrated.
10. No sensitivity or uncertainty analysis is presented for the GM(1,1) predictions, logistic regression coefficients, or scenario assumptions.

Constructive feedback

The reference list appears adequate but could include more recent studies on grey models, CLUMondo applications, and arid-region land-use simulation. Some foundational LUCC modeling literature is cited, but more recent machine-learning–based approaches could be acknowledged. The manuscript should explicitly discuss limitations related to data quality, model assumptions, and scenario uncertainty. The absence of climate-change projections is a notable omission for an arid-region study. The methods section should be reorganized to separate data sources, preprocessing, model construction, and scenario design more clearly. The results section would benefit from more quantitative comparisons across scenarios, including spatial metrics and cropland-loss statistics. Please provide full parameter tables for GM(1,1), logistic regression, and CLUMondo settings. Try to include validation plots, ROC curves, or confusion matrices to support accuracy claims. You must improve figure readability and ensure all symbols and abbreviations are defined.

Methodological issues

The improved GM(1,1) model lacks a rigorous validation procedure. The manuscript does not compare it with baseline models. The CLUMondo transition rules and elasticity settings are not fully documented, limiting reproducibility.Logistic regression suitability modeling requires diagnostics (VIF, ROC curves, residual analysis). These are not reported. AUC values above 0.8 are mentioned, but no class-specific performance metrics are provided.  The assumption that past land-use trends can be extrapolated to 2040 may not hold under rapid policy or climate changes.  The study assumes stable socioeconomic trajectories without discussing uncertainty ranges.

The manuscript does not address potential biases in land-use survey data or inconsistencies between different data sources. Downscaling socioeconomic data introduces uncertainty that is not quantified. Objectives are clearly stated, but the conclusions sometimes overstate the certainty of the results. Some typographical errors and formatting inconsistencies remain.

 

Summary:

Some sections are overly long and contain unnecessary narrative detail. The scientific writing requires refinement to improve precision and readability. Figures, tables, and model descriptions need clearer explanations and stronger alignment with the stated objectives.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have revised the manuscript according to my suggestions. I would like to see it published.

Author Response

Comment1: The authors have revised the manuscript according to my suggestions. I would like to see it published.

Response:Thank you for your positive assessment and for the constructive suggestions in the first round. We have revised the manuscript accordingly and improved clarity and consistency throughout. We appreciate your support for publication.

Reviewer 2 Report

Comments and Suggestions for Authors

The authors made good efforts to revise the paper, and indeed the quality improved.

Author Response

Thank you for your positive feedback. We are pleased that the revised manuscript is of higher quality. In this revision, we further enhanced the clarity and consistency of the methods and results sections and improved the presentation of tables and figures. Thank you for your time and support.

Reviewer 3 Report

Comments and Suggestions for Authors

• The introduction has been updated with recent studies on LUCC/cropland scenario simulations in China and now includes “research questions”, although these are phrased as statements rather than actual questions. However, a clearer distinction is still needed between the methodological contribution (e.g., the “improved” demand module) and the application of the established CLUMondo framework.
• The authors have added descriptions of model training (sliding window), evaluation metrics (MAE/RMSE/MAPE), and defined a benchmark design. However, the same issue persists in the results: the “Benchmark comparison” (Table 7) presents only baseline methods and does not clearly report the Improved GM(1,1) result as a comparable row. Thus, there is still no quantitative evidence supporting the claim that the improved model performs better. Furthermore, the table shows that the persistence baseline performs best in terms of MAE/RMSE/MAPE, which contradicts the “improved precision” unless this is explicitly explained.
• Inconsistency in accuracy metrics (Kappa/OA) has been partially resolved, but potential confusion remains:

In version 1, the reported values were contradictory (e.g., different OA/Kappa values in the text and the table).

In version 2, the numbers in the abstract and conclusion were corrected, but Kappa = 0.87 now appears as the elasticity calibration outcome (in the text near Table 3), while Table 6 still reports Kappa around 0.85 for validation. This needs to be clearly disentangled (calibration vs. validation; which year or time period; which metric is being reported).

• The issue with figures (readability of legends, font sizes, and print/PDF layout) has not been addressed consistently across all figures.

Comments on the Quality of English Language

The English could be improved, but I am not a native speaker, so I can't assess its quality in full.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

The authors have substantially improved the contents, methodological rigor, and overall structure of the paper. Their revisions strengthened the scientific narrative, enhanced clarity across sections, and contributed to a more coherent and impactful manuscript. I appreciate the considerable effort invested in these improvements.

Author Response

评论：作者们对论文的内容、方法论的严谨性和整体结构都进行了大幅改进。他们的修改增强了科学叙述的力度，提高了各章节之间的清晰度，并使论文更加连贯有力、更具影响力。我非常感谢他们为这些改进所付出的巨大努力。

回复：非常感谢您给予的积极反馈和详尽审阅。我们很高兴您对我们为加强内容、方法严谨性、结构和整体连贯性而进行的诸多修改表示认可。我们衷心感谢您抽出宝贵时间并给予我们支持。