Dynamics of Cropland Non-Agriculturalization in Shaanxi Province of China and Its Attribution Using a Machine Learning Approach
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsCropland non-agriculturalization(CLNA) has globally occurred in recent decades, particularly in rapidly developing countries like China. The manuscript integrates multiple methods, including machine learning, to analyze the spatial clustering patterns of CLNA in Shaanxi Province and quantify the contributions of various driving factors. I think the overall logic is clear and results are reasonable, but the following issues can be improved:
1. The study area introduction lacks alignment with the research objectives. It should emphasize recent trends in CLNA, key socioeconomic and environmental changes in Shaanxi Province, and its representativeness within China and globally to express why the topic is important.
2. The manuscript adopts multi-methods like Geo-detector and XGBoost-SHAP, but lacks a detailed comparison of these methods, their selection rationale, and how they address limitations of existing approaches. Also, key model parameters need to insufficiently described, affecting reproducibility. It is very important.
3. The selection criteria and sources of driving factors are unclear. The results section should provide mechanistic explanations for significant factors, linking them to the study area's characteristics. Without this, the results risk being perceived as "data in, data out."
4. The discussion and conclusion sections are is so simplistic that I doubt whether the authors have truly reflected on the significance of the research. Obviously, these two sections lack in-depth analysis, comparison with existing studies, practical recommendations for cropland protection, and so on. The conclusion focuses on results excessively but fails to summarize methodological contributions or highlight innovations.
5. Some formatting and clarity issues: overly lengthy method descriptions, insufficient figure captions like Fig. 9, and outdated or incomplete references like 4 and 8.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis paper analyzed spatiotemporal dynamics and driving factors and their interactions of CLNA in Shaanxi Province. Results showed that XGBoost was the best ML method to predict CLNA and precipitation was the most important fact in the process of cropland non-agriculturalization, and the interaction effects between these 15 factors contributed more than individual factors. Although the subject of the paper, analysis of CLNA spatiotemporal dynamics, is not necessarily a novelty, it could be one more contribution to the research field of LUCC. The paper was well organized and the results were clearly illustrated. However, I have some concerns which need to be addressed during revision. I list my comments as follows for further improving:
1. Introduction section, I suggest authors to pay more attention on the LUCC researches that conducted on the region of SP and their shortcomings to highlight your work.
2. Lines 113-117, you outlined the meteorological characteristics, please give the references.
3. Line 154, Xi and Xj present the non-cropland area, or the area of cropland converted to non-cropland (CLNA)?
4. Line 191, about the calculation of q, I suggest to cite the original paper: Wang JF, Zhang TL, Fu BJ. 2016. A measure of spatial stratified heterogeneity. Ecological Indicators 67: 250-256.
5. Line 217, how to divide the data into the training and testing sets, randomly?
6. Please give the equations to calculate RMSE and MAE, and corresponding units.
7. Figure 3, I don’t think the lines introduced from linear regression can bring valuable information here. Readers can find the trend easily from the bar chart.
8. Line 250, “the extent of non-cropland…… with an increasing magnitude of 7506.4 km²……”, this is not consistent with “14033.1 km² of cropland was converted to non-cropland” in abstract.
9. Line 275, “greater than 5”, it should be 0.5 according to Table 1.
10. Section 3.3, authors employed different models to predict CLNA and compared their metrics, but did not give the details of parameters optimization which may has a significant impact on model’s performance, such as RF which were widely used in classification and regression tasks.
11. Table 2, since the training data has been used for training models, it is not necessary to compute R2, RMSE, and MAE using training set. In my opinion, only the metrics derived from the data which was not involved when training a model can make sense.
12. Discussion section, I suggested to the authors to make the discussion of the results based on some more related bibliography - specially pay more attention on the attribution analysis.
13. Figure 9, panel (a), Geo-detector gives the definition of q-value and it can be easily calculated, but I am confused on how to get the q-value from XGBoost?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsPlease see the attcahed report
Comments for author File: Comments.pdf
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe manuscript has been revised to address the issues raised, making the overall content clearer, and some detailed problems have also been improved. However, the current revision still fails to clarify my key concern: what exactly is the purpose and significance of this study? I think the depth of background information is insufficient, the regional interpretability is lacking, and how the results can be applied remains unclear.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsMy concerns were addressed, and the manuscript was significantly improved.
Author Response
Thank you once again for your time, as well as your insightful comments and suggestions, which have significantly improved the quality of the manuscript.
Reviewer 3 Report
Comments and Suggestions for AuthorsThe authors addressed all of my comments
Author Response
Thank you once again for your time, as well as your insightful comments and suggestions, which have significantly improved the quality of the manuscript.
