Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors This paper focuses on the impact of complex mountainous terrain on the accuracy of Land Use Products (LUPs), systematically evaluates the applicability of 8 mainstream LUPs in the Wumeng Mountain area, and quantifies the influence mechanisms of topographic factors (such as elevation and surface complexity) by combining XGBoost and SHAP models. The topic of this paper is of practical significance, with a feasible methodological framework and reasonable results. However, there are still some aspects that need improvement:

1. The generation process of the reference data needs to be supplemented. What kind of images were used? When were the images taken? Whether there is an impact of temporal dynamics on the accuracy of LUPs. Meanwhile, details of quality control during the vectorization process should be added to support its reliability as the "true value".
2. In Table 1, the overall accuracy of "ESA_CCI" is marked as "73.00" (lacking the unit %).
3. In Table 2 and Figure 6, what does MV_LUCC stand for? As it is abbreviated for the first time, the full name should be supplemented.
4. In Figure 1b, the font is too small to be clearly seen.
5. There is a contradiction between "the average accuracy curve exceeds 0.8 above 3500 m" in Figure 9 and "the overall accuracy (OA) of all products is below 0.7" in the text. It is necessary to verify the data calculation logic (whether it refers to the accuracy of specific land types rather than the overall OA).
6. Consistency of terms: After "Land Use Products" appears for the first time, it should be uniformly abbreviated as "LUPs" to avoid alternating use throughout the paper.
7. The parameters of the XGBoost model (such as tree depth, learning rate, number of iterations) are not specified, which affects the reproducibility of the results. Details of model training need to be supplemented.
8. Why the accuracy increases with elevation? This result needs further explanation.
9. It is necessary to supplement the analysis and discussion on the similarities and differences with the results of similar mountain assessment studies (such as reference [21]), and compare the differences in the conclusions of this study.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The study evaluates the accuracy and applicability of eight global land use/land cover (LUC) products in a representative mountainous region of China (Wumeng Mountains), where complex terrain complicates remote sensing classification. Using manually vectorized high-resolution images as a reference, metrics such as overall accuracy (OA), area deviation rate (ADR), and confusion matrices were applied to compare the consistency and performance of each product. The topic is worthy of research; however, it presents several minor changes that need to be addressed before it can be considered for publication.

General comments

C1. The main concern is that the novelty of the research is not fully clear. If such novelty is not clearly highlighted, the risk is that the manuscript looks more a simple case study rather than a research paper.

C2. What are the major contributions of this study? should be carefully mentioned in the discussion section.

*The answer to these questions should be reflected in the manuscript.

Specific comments

Lines 36-39: The authors should rephrase the sentence to make the causal relationship between land use change and ecosystem impact explicit.

Lines 45-50: The justification for the importance of mountain regions is solid, but it would be useful to cite recent studies to reinforce the current relevance of the topic.

Lines 61-67: Mention that the lack of standardized validation protocols is a recognized limitation in the literature (include sources), and could be enriched by citing initiatives that have sought to harmonize (e.g., Land Cover Classification System, FAO, or others).

Line 85: How does this study specifically address these limitations?

Line 92: Add a sentence connecting the ideas before the research questions to emphasize the knowledge gap.

Line 116-117: The term “highly heterogeneous” should be quantified with some topographic metric (e.g., coefficient of variation).

Line 119: According to whom? Support with relevant literature.

Line 121: Figure 1c. One would expect to see a map with these distributions or classifications.

Line 129: Why were these 8 specific products selected, what were the inclusion/exclusion criteria?

Line 142: Change “The” to “the.”

Line 142: It is important to mention the “Modifiable Area Unit Problem” (MAUP) and it would be valuable to acknowledge this inherent limitation when comparing products with such disparate resolutions.

Line 153: Detail the visual interpretation process (e.g., number of interpreters, consensus criteria to reduce subjectivity, which combinations were taken into account, textures were taken into account). Mention potential limitations (e.g., human errors in data entry).

Line 165–168: The inclusion of shrubland in the forest class is common, but it is important to emphasize the risk of overestimation. How could the risk of overestimation be addressed?

Line 172-174: The quality control process is adequate, but it would be advisable to mention what type of cross-validation was applied or whether there was peer review.

Line 176: A space is missing.

Line 177: Increase the size of the figure.

Line 212-239: Although the study uses conventional metrics (PA, UA, F1, and OA), the authors should include or contrast with the methodological framework of Olofsson et al. (2013, 2014, https://doi.org/10.1016/j.rse.2012.10.031, https://doi.org/10.1016/j.rse.2014.02.015), which allows for the estimation of area-adjusted errors, confidence intervals, and weights by sampling design. This would be especially valuable in mountainous contexts where classes are not distributed homogeneously and there is high spatial fragmentation.

Line 258: Why was XGBoost chosen over other models (e.g., Random Forest), given its use in similar studies?

Line 261-263: The authors should explicitly mention whether data were divided into training and testing, and what fit metric was used (e.g., RMSE, R²). They should also indicate how many trees were used or whether there was cross-validation.

Line 278-279: Is this due to spatial resolution or classification schemes?

Line 295: The authors should add confidence intervals or significance tests to validate differences between products.

Line 297: The figure is not clear, increase your size. The same applies to all figures.

Line 338-339: This finding is consistent with recent literature. Cite works that agree.

Line 341: Explicitly suggest that LUPs should be combined with auxiliary data in mountainous areas (DEM, SAR).

Lines 423-425: The DynamicWorld pattern is interesting. It is recommended to explore further whether this is due to its time window (frequent updating) or its algorithm (FCNN). This could strengthen the discussion.

Lines 468-471: Propose specific strategies for these areas in the conclusions.

Line 472: The “lack of topographic correction” is mentioned. Here, the implementation of this process could be suggested.

Lines 479-481: There is mention of “extrapolation bias,” but it would be interesting to quantify it or suggest some training metric.

Lines 575-577: The proposal for SAR-optical fusion in shaded areas is appropriate, but it could be mentioned that this technique still faces adoption issues.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The research evaluates the applicability and performance of eight land use products (LUPs) with different spatial resolutions, ranging from 1410 m to 1000 m, in a typical mountainous area of the Wumeng Mountains. Manually vectorized high-resolution imagery was used as reference data. While the study does not provide a major scientific contribution to remote sensing methodologies, the topic is relevant to land use mapping and offers a valuable practical application in this field. The study area is relatively small, which limits direct comparison with global datasets; however, the evaluation of several land use products enhances its significance compared with other published work.

Regarding the research methodology, it would be more appropriate if the authors had used Landsat data to classify land use and land cover features in the study area and then compared these classifications with the land use products. Such an analysis could strengthen the study, particularly given the small scale of the research area. The conclusion is presented as bullet points rather than a coherent discussion. It should instead summarize the main research questions, highlight the key findings, and provide connected insights along with recommendations for future work.

The references used in the study are appropriate.

I also concern about the research structure: Several sentences and paragraphs are overly long, which affects readability. For example, the second paragraph in the introduction is too lengthy; I recommend dividing it into two shorter paragraphs and restructuring the sentences to follow a logical sequence (e.g., using “first,” “second,” “third,” etc.). Additionally, the sentence in lines 157–161 is excessively long and should be broken into shorter, clearer statements. At lines 218, 224, and 237, new paragraphs should be introduced for better flow and clarity.

Regarding Table 3, the header currently reads: “Extracted using Geomorphon algorithm in r.geomorphon or similar module.” Please clarify whether the r.geomorphon module was actually used in this study or whether another module was applied. If another module was used, specify it; if not, remove “or similar module,” as the table header should reflect only the factors actually applied in the research. Please clarify whether the links cited in the manuscript must be presented directly in the text or should instead be moved to the references section.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf