Review Reports - The Impact of Ecological Restoration Measures on Carbon Storage: Spatio-Temporal Dynamics and Driving Mechanisms in Karst Desertification Control

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper took Zhijin County as the research object, aiming to guide regional vegetation construction and has theoretical and practical significance. However, this study has some serious problems.

(1) The karst landform and habitat are unique, but the carbon density assessment model is not unique. At least, it is not obvious. Please list the assessment model in detail. Which factors were considered?

(2) The ecological engineering is not detailed enough. An ecological project of a county should list the main ecological projects, such as the number of acres of reforestation, including the number of tree forests and grasslands?

(3) Please list the Date sources in a table. The table mainly included name, year, resolution, source, etc.

(4) Is Figure 3 a problem of representation or writing? Why are there only three data in the legend?

(5) In Figure 4, what is constant? What is variable? The author should be very clear. Why are 2020 and 2025 so large? And these two years are not exactly same? Why does 2030 become smaller? Why does 2010 and 2015 also become larger? I don't know what it means. The composition of the ecosystem should be the same.

(6) The classification method in 3.2 needs to be unified. In the current classification methods of artificial forests, forest closure, and other regions, should forest closure include artificial forests? what is this different from the previous artificial forests? Please explain in detail.

(7) In the driver analysis, why are the common factors used? Why are there no unique factors at the county scale? Here, it should be the driver analysis of county-specific factors.

(8) This paper is a carbon density analysis of all land types. The title is also about the impact of forest enclosures? It seems a bit off. Forest is the main body, and it does not account for half of the composition.

(9) Discuss the following issues. Each section should be one paragraph. Why are there not two or three paragraphs? Each section does not show the elevated content, lacks investigation, lacks specific problem-specific analysis, and it is not clear about the content of county-scale analysis. The analysis also has problems. The west is high and the east is low. This should be related to the distribution of landforms, ecosystems, and settlement distribution, but the author talks about the hydrological process. There is also a matter related to literature. The previously mentioned literature should not be repeated later.

In conclusion, this article, like other large-scale studies, lacks the necessary research at the county level and fails to provide specific analyses at the township and village analyses. It fails to reveal the unique characteristics of a county. Therefore, I personally suggest that this article be revised significantly before being reviewed again. In this section on cold and hotspots analysis, it should combine the actual ecological projects and construction of the county, and propose targeted ecological construction goals, priorities, and requirements, rather than conducting a macroscopic analysis as is done in large-scale studies.

Author Response

Comments 1: [The karst landform and habitat are unique, but the carbon density assessment model is not unique. At least, it is not obvious. Please list the assessment model in detail. Which factors were considered?]

Response 1: Thank you for your key comments. A table of model-specific assessment parameters has been added to the article (section 2.4.2). The focus of this study is to solve the problem of timeliness of dynamic carbon intensity assessment. Regarding the establishment of the simulation, the most important consideration is the selection of data points (including data points of AGC, BGC, SOC and DMC). Regarding the selection of data points is obtained by fitting and comparing the R² (mainly comparing the AGC): 1. using the data points of the same latitude in Guizhou province, this set of data points includes the points of the same latitude globally, totaling 629 points, and fitting to obtain the R² = 0.586 for this model; 2. using the data points of Guizhou, Guangxi, and Yunnan according to the data points used in the paper of [1], totaling 401 data points, and fitting to obtain the R² = 0.602; 3. using only data points from Guizhou, with a total of 93 data points, the fitted R² = 0.642. Comparison reveals that the highest R² is obtained by using data points from Guizhou province, so this data point is chosen for fitting. The second reason for selecting data points from Guizhou province was that, given the spatial heterogeneity of habitat factors in karst areas, Guizhou province, as the province with the widest coverage of karst areas in China, retained the uniqueness of the carbon density data points in karst areas, and the overall environment was similar to that of Zijin County, which was crucial to the study of the carbon density in Zijin County.

Both AGC and BGC use dynamic carbon density, while SOC and DNC use static carbon density. Specifically, the carbon density values for AGC and BGC were calculated using the regression equations for the corresponding years, taking the carbon density values for each land class in Representative 1. Instead of DMC and SOC, static carbon density values were used directly due to poor fitting (Section 2.4.2). This limitation is clearly stated in the discussion section (Section 4.3).

Table 1. Model parameter evaluation.

Carbon density	R²	Coefficient	Value	SE	t	p
AGC	0.642	Intercept	15.626	3.1234	5.0024	<0.01
AGC	0.642	Slope	0.838	0.066	12.7731	<0.01
BGC	0.364	Intercept	4.575	1.0131	6.9403	<0.01
BGC	0.364	Slope	0.223	0.0322	8.6266	<0.01
DMC	0.018	Intercept	1.413	0.5271	0.0406	<0.05
DMC	0.018	Slope	0.043	0.0052	0.8558	–
SOC	0.001	Intercept	149.77	144.95	15.97	<0.01
SOC	0.001	Slope	0.332	0.2849	1.1543	–

1.Luo, D., Zhou, Z., Zhang, L., et al. Evolution and driver analysis of forest carbon stocks in karst mountainous areas of southwest China in the context of rocky desertification management. Catena 2024, 246, 108335.

The ecological engineering is not detailed enough. An ecological project of a county should list the main ecological projects, such as the number of acres of reforestation, including the number of tree forests and grasslands?

Response 2: Thank you for your valuable guidance. We have now added the area of newly planted trees for afforestation (2009–2019) and forest closure (2006–2019), and revised this section accordingly. For these two ecological restoration projects, we have included additional descriptions of their implementation characteristics and methods. Since the dataset does not include a field for “forests and grasslands,” this category has not been added. Instead, we use tree canopy closure and shrub cover to characterize forests and shrublands (including grasslands). Regarding whether forest closure should include afforestation, these are two distinct technical approaches, with the core distinction lying in the dominant restoration force and intervention intensity. Forest closure primarily relies on natural restoration, with artificial intervention limited to auxiliary measures (such as replanting). Although the replanting ratio is high (57.91%), replanting is a localized intervention rather than comprehensive afforestation, and the tree species used are water-conserving trees such as birch and cypress, which can still be considered as artificially promoted forest closure. Afforestation, on the other hand, involves high-intensity artificial intervention, which conflicts with the principles of forest closure. It primarily uses fast-growing economic tree species such as soapberry and pomegranate. The tree species planted are different, and the modes are also different (afforestation is comprehensive afforestation, while forest closure is spot replanting). Therefore, forest closure does not include afforestation.

Comments 2: [The ecological engineering is not detailed enough. An ecological project of a county should list the main ecological projects, such as the number of acres of reforestation, including the number of tree forests and grasslands?]

Comments 3: [Please list the Date sources in a table. The table mainly included name, year, resolution, source, etc.]

Response 3: Thank you for your attention to detail in the article. We have now revised this section (Section 2.3) according to your suggestions and created a new table that clearly lists the data names, resolutions, and sources. We have also added a description of the data preprocessing. The forest inventory data and desertification data are classified as confidential information. All such data in the article has been de-identified to ensure it does not contain any classified information related to terrain or other sensitive details, and is solely used for the purpose of writing this article. The data year has not been included in the table; instead, it is described in the text. Furthermore, corrections have been made to data source websites that were previously inaccessible, ensuring that the websites are now accessible.

Comments 4: [Is Figure 3 a problem of representation or writing? Why are there only three data in the legend?]

Response 4:Thank you for your question. Regarding Figure 3, I have reclassified it into three categories to make it more intuitive. This is because if the original grid data is stretched and displayed directly, it is difficult to capture spatial changes, and it is difficult to observe with the naked eye. Therefore, I used the reclassified image for display.

Comments 5: [In Figure 4, what is constant? What is variable? The author should be very clear. Why are 2020 and 2025 so large? And these two years are not exactly same? Why does 2030 become smaller? Why does 2010 and 2015 also become larger? I don't know what it means. The composition of the ecosystem should be the same.]

Response 5: Thank you for raising this question, which is crucial to the rigor of this paper. Figure 4 represents changes in land use area. I have added a table showing changes in area (see table below), which shows that the area of each land use type has changed, so there is no constant value. The land use changes for 2025 and 2030 were simulated and predicted using the CA-Markov model, with the software IDRISI Selva. The Kappa value is 0.88, indicating that the prediction results are satisfactory and meet the experimental requirements (Kappa > 0.75). The Sankey diagram for 2015–2020 shows significant flow, reflecting the intense land use changes observed during this period. According to the data, during this phase, forest land decreased by 1,808.44 ha (-0.8%), while construction land increased by 563.95 ha (+11.2%). From 2020 to 2025, significant changes in cropland, woodland, and construction land led to large flows in the Sankey diagram. The operational logic of the CA-Markov model in IDRISI is as follows: when the Markov chain calculates transition probabilities based on historical change rates (2010–2020), if a particular transition (e.g., cropland converting to grassland) suddenly accelerates in the short term (2015–2020), the model will exponentially amplify this trend. The spatial allocation of cellular automata (CA) further focuses on these changes, leading to a “sudden mutation” in 2025. However, the changes from 2015 to 2020 are based on real statistical data, and the model's predicted 2025 mutation is an amplified effect of the short-term acceleration of land use changes, while the actual observed land use changes from 2015 to 2020 are a continuation of historical trends.

Comments 6: [The classification method in 3.2 needs to be unified. In the current classification methods of artificial forests, forest closure, and other regions, should forest closure include artificial forests? what is this different from the previous artificial forests? Please explain in detail.]

Response 6:Thank you for your valuable feedback. I have added a clearer classification in Section 3.2. In this study, we adopted an exclusive classification system: artificial forests refer to forest land that has been intentionally planted and managed through human intervention; closed forests refer to natural secondary forests that are mainly restored through natural regeneration processes; other areas refer to all areas not included in the above two categories, including grasslands, shrublands, water bodies, agricultural land, and urban areas. We maintain these two categories for independent analysis because: this directly validates our core hypothesis comparing artificial intervention with natural restoration patterns; the ecological processes and carbon sequestration mechanisms of the two categories differ fundamentally, and they provide different policy implications for ecological policies targeting different restoration patterns.
Regarding whether forest closure should include afforestation, these are two distinct technical approaches, with the core distinction lying in the dominant restoration force and intervention intensity. Forest closure primarily relies on natural recovery, with artificial intervention limited to auxiliary measures (such as replanting). Although the replanting proportion is relatively high (57.91%), replanting is a localized intervention rather than comprehensive afforestation, and the tree species used are water-conserving trees such as birch and cypress, which can still be considered as artificially promoted forest closure. Afforestation, on the other hand, involves high-intensity artificial intervention, which conflicts with the principles of forest closure. It primarily uses fast-growing economic tree species such as soapberry and pomegranate. The tree species planted are different, and the modes are also different (planted forests involve comprehensive afforestation, while forest closure involves spot replanting). Therefore, forest closure does not include planted forests.

Comments 7: [In the driver analysis, why are the common factors used? Why are there no unique factors at the county scale? Here, it should be the driver analysis of county-specific factors.]

Response 7:

Thank you for your valuable feedback. The analysis of the correlation in the article was calculated using Pearson's correlation coefficient. This is because the two main subjects, artificial afforestation and forest closure, have a very large dataset, with hundreds of thousands of data points. Using Pearson's correlation coefficient allows us to include every data point, resulting in a more accurate correlation. Regarding the point that there is no single factor for counties, this is because the patches for artificial afforestation and forest closure are highly fragmented (as seen when opened in ArcGIS 10.8, refer to Figure 1) and distributed extremely unevenly. Therefore, the study separately calculated the correlations between artificial afforestation, forest closure, and NEE. Since each patch differs in terms of geographical conditions, climatic conditions, and human impacts, the correlations between carbon storage and these factors also vary, resulting in no single dominant factor. If a direct correlation analysis of carbon density in Zhenjin County were conducted, it would weaken the article's analysis of the differences in artificial afforestation and forest closure, leading to a superficial analysis akin to those at larger scales. According to your suggestion, using Pearson's correlation coefficient directly in the analysis would overlook the effects of interactions between factors. To address this, we considered two methods for interaction analysis. The first was modeling, but due to the numerous influencing factors, the resulting equations and model parameters became too complex, and the modeling process was too challenging, so this method was not adopted. The second method involves using a geographic detector to analyze factor correlations and interaction effects. First, we convert the original data into point data (hundreds of thousands of points) and perform random sampling or stratified sampling to obtain 1,000 sample points (sampling is necessary because the geographic detector cannot process hundreds of thousands of data points, and 1,000 data points meet the research requirements). Alternatively, the fishing net tool was used to extract points within the fishing net for spatial connection (2,000–6,000 sample points). When performing sampling or using the fishing net tool, we reclassify most of the data into 9 categories based on natural breakpoints, while some data, such as desertification, are classified into 5 categories. Therefore, we classify these already categorized data with fewer categories into 4–5 categories. After importing the various processing results into the Geographic Detector, the correlations remain low. Even after multiple verifications and repeated operations, the correlations and interaction effects still fail to meet the research requirements. Therefore, we choose to retain the original correlation results (Pearson correlation coefficient). The Geographic Explorer results are shown as follows:

Comments 8: [This paper is a carbon density analysis of all land types. The title is also about the impact of forest enclosures? It seems a bit off. Forest is the main body, and it does not account for half of the composition.]

Response 8:Thank you for your valuable feedback. We have fully incorporated your suggestions. The article calculates carbon storage using land-use change and carbon density predictions to derive carbon density. As you pointed out, the main focus of the article was unclear. We have now expanded the depth of the writing based on your feedback, adding a new section (Section 3.3) to the results section. This section specifically addresses the differences in carbon density between karst desertification areas and non-karst desertification areas, as well as between planted forests and forest closure. The article is written from the unique perspective of Zhenjin County's karst desertification, highlighting the core content of the article—the differences between planted forests and forest closure—while also delving deeper into karst regions. Additionally, the title of the article has been revised to better align with its core content and the characteristics of karst desertification. The revised title is: “The impact of ecological restoration measures on carbon storage: Spatio-temporal dynamics and driving mechanisms in karst desertification control.”

Comments 9: [Discuss the following issues. Each section should be one paragraph. Why are there not two or three paragraphs? Each section does not show the elevated content, lacks investigation, lacks specific problem-specific analysis, and it is not clear about the content of county-scale analysis. The analysis also has problems. The west is high and the east is low. This should be related to the distribution of landforms, ecosystems, and settlement distribution, but the author talks about the hydrological process. There is also a matter related to literature. The previously mentioned literature should not be repeated later.]

Response 9:Thank you for your valuable review comments. We have incorporated your suggestions by adding summaries to each subsection to highlight the key points and conducting further investigations and analyses on planted forests and forest closure. Regarding the issue of unclear county-level analysis content, we have added a section (Section 3.3 of the revised paper, “Differences in Carbon Density Under Different Levels of Rocky Desertification”) to further analyze the carbon density differences of planted forests and forest closure in conjunction with the actual situation in Zhenjin County (rocky desertification), thereby enhancing the expression of the article's main content and making it more prominent. We have also revised the title to better align with the original text. Regarding the hydrological processes, the article indeed did not provide a more in-depth or detailed description, only a brief mention. After discussion, we decided to remove this section as it was not strongly related to the main content of the article. Regarding the issue of literature duplication, the article did indeed have this issue, and we have adjusted the literature to meet the journal's requirements. We have also simplified the abstract section of the study.

Reviewer 2 Report

Comments and Suggestions for Authors

It is important for the authors to consider the following points:

Limitations in the calibration of the InVEST model: AGC and BGC values were appropriately adjusted; however, DMC and SOC were not corrected, which may result in an underestimate of total carbon stocks. Although this is explained in the study's limitations, it is essential to clarify how these limitations affect the results.
The ambiguity in the definition of specific terms needs to be clarified, especially terms like "closed forest," which should be more precisely operationally defined to facilitate replicability.
It is essential to explain the reasons behind the drivers selected to account for changes in carbon stocks.
Other ecosystem services that ecological restoration could provide need to be analyzed, as well as what might increase the costs and benefits of closed forests, for example, social and economic participation.
A distinction needs to be made between ecological restoration and the landscape restoration used in the research. They are two different concepts that are not clear.

Lines 20-21: The carbon decline between 2020 and 2025 deserves further contextual discussion. Were there specific public policies or climate events that explain this?
Lines 243-255 (Table 1): It would be helpful to include confidence intervals or standard errors for the estimated carbon densities.
Line 346 (Table 3): It is advisable to show the number of hectares involved by category, in addition to growth rates, to get a better idea of the spatial impact.
Line 374 (Table 4): Although many variables are reported, it is unclear whether a multivariate analysis or only bivariate correlations were performed. This can lead to misinterpretations due to multicollinearity.
Figure 5: The legend could be clearer: it is currently difficult to distinguish the carbon density curves for each restoration type.
Line 517-520: A more complete correction of the InVEST model is recommended, or at least to discuss the implications of not adjusting SOC and DMC in the limitations section.

Author Response

Comments 1: [Limitations in the calibration of the InVEST model: AGC and BGC values were appropriately adjusted; however, DMC and SOC were not corrected, which may result in an underestimate of total carbon stocks. Although this is explained in the study's limitations, it is essential to clarify how these limitations affect the results.]

Response 1:Thank you for your attention to the overall rigor of the article. Regarding the impact of these limitations on the results, the discussion section in 4.3 does indeed provide a somewhat vague description. The soil carbon cycle has a duration of 0.9–152 years, with an average of 24.3 years, which is a relatively long cycle [1]. However, the forest closure (2006–2019) and afforestation (2009–2019) periods used in the study do not reach the average soil carbon cycle duration of 24.3 years. Additionally, both forest closure (2006–2019) and afforestation (2009–2019) involve the annual addition of new trees. Forest closure primarily relies on natural recovery, with human intervention limited to auxiliary measures (such as high proportions of spot planting at 57.91%), while afforestation involves comprehensive planting. The soil carbon density cycles of these newly added trees are shorter than those of the previous year’s trees. The study concludes that changes in soil carbon density during the implementation phase can be neglected [2]. Regarding DMC, the forest litter layer exhibits strong respiratory activity, with rapid carbon turnover rates leading to weaker carbon sequestration capacity. Therefore, the carbon density of the litter layer can be considered a constant value [3]. Since SOC and DMC are not significantly different, no corrections were applied. This limitation has a limited impact on the study. This section has been revised (Section 4.3).

1.Wang, J., Sun, J., Xia, J., He, N., Li, M., & Niu, S. (2018). Soil and vegetation carbon turnover times from tropical to boreal forests. Functional Ecology, 32(1), 71-82.

2.Luo, D., Zhou, Z., Zhang, L., Chen, Q., Huang, D., Feng, Q., ... & Wu, L. (2024). Evolution and driver analysis of forest carbon stocks in karst mountainous areas of southwest China in the context of rocky desertification management. Catena, 246, 108335.

3.Carvalhais, N., Forkel, M., Khomik, M., Bellarby, J., Jung, M., Migliavacca, M., ... & Reichstein, M. (2014). Global covariation of carbon turnover times with climate in terrestrial ecosystems. Nature, 514(7521), 213-217.

Comments 2: [The ambiguity in the definition of specific terms needs to be clarified, especially terms like "closed forest," which should be more precisely operationally defined to facilitate replicability.]

Response 2:I have added a clearer classification in Section 3.2. In this study, we adopted a mutually exclusive classification system. Planted forests refer to forest land that has been intentionally planted and managed through human intervention; forest closure refers to natural secondary forests that are restored mainly through natural regeneration processes; other areas refer to all areas not included in the above two categories, including grasslands, shrublands, water bodies, agricultural land, and urban areas. We maintain these two categories for independent analysis because: this directly validates our core hypothesis comparing artificial intervention with natural restoration patterns; the ecological processes and carbon sequestration mechanisms of the two types of areas differ fundamentally, and they have different implications for ecological policies regarding different restoration patterns.

Comments 3: [It is essential to explain the reasons behind the drivers selected to account for changes in carbon stocks.]

Response 3:Thank you for your professional opinion, which is very important to the overall quality of the article. In sections 3.1 and 3.2 of the article, we did not clearly select the corresponding driving factors for in-depth analysis. Since carbon stock calculations are based on carbon density tables and land use changes, the article analyzed land use changes but did not analyze in depth the changes in carbon density of various factors in the carbon density tables. We have now incorporated your suggestions and conducted a more in-depth analysis of the causes of carbon stock changes in the relevant results and analysis sections, while also expanding the discussion on how influencing factors impact carbon stock changes.

Comments 4: [Other ecosystem services that ecological restoration could provide need to be analyzed, as well as what might increase the costs and benefits of closed forests, for example, social and economic participation.]

Response 4:Thank you for your constructive feedback. We are considering adding a subsection to the Results and Analysis section to specifically address the differences in carbon density between areas with varying degrees of karstification and non-karstified regions, as well as between planted forests and forest closure. This subsection will be written based on the unique karstification characteristics of Zhenjin County, highlighting the main focus of the article—the differences between planted forests and forest closure—while also deepening the article's exploration of karst regions. We will also revise the title of the article to better align with its main content and the characteristics of desertification. Adding this subsection will help analyze the comprehensive benefits of the two ecological restoration measures on carbon storage in karst regions. Selecting different ecological restoration measures based on varying degrees of desertification will contribute to mitigating desertification in karst regions.

Comments 5: [A distinction needs to be made between ecological restoration and the landscape restoration used in the research. They are two different concepts that are not clear.]

Response 5:Thank you for your professional insights. Your attention to detail in the article has indeed enhanced its overall coherence and logical structure. We agree that clarifying these concepts is crucial for accurately constructing our research framework. In our study, the measures implemented (i.e., planted forests and forest closure) clearly fall under the category of ecological restoration rather than landscape restoration. This is because, as required by regional ecological engineering, their primary objective is to restore ecosystem functions and services, particularly addressing soil erosion and enhancing carbon sequestration—a critical ecosystem service—rather than merely improving visual aesthetics or land cover. To avoid any potential misunderstandings, we have now clearly defined the concept of “ecological restoration” used in this paper in Section 2.2 (see lines 2–3 in Section 2.2 of the revised draft). We emphasize that our work focuses on assessing the restoration of ecosystem functions (carbon storage), not landscape appearance. Forest closure and afforestation are two distinct technical approaches, with their core distinction lying in the dominant restoration force and intervention intensity. Forest closure is primarily based on natural restoration, with artificial intervention limited to auxiliary measures (such as replanting). Although the replanting ratio is relatively high (57.91%), replanting is a localized intervention rather than comprehensive afforestation, and the tree species used are water-conserving trees such as white birch and cypress, which can still be considered as artificially promoted forest closure. Afforestation, on the other hand, involves high-intensity artificial intervention, conflicting with the principles of forest closure. It primarily uses fast-growing economic tree species such as soapberry and pomegranate. The tree species planted differ, and the modes are also distinct (afforestation involves comprehensive afforestation, while forest closure involves spot replanting). The subjects of the study: planted forests and forest closure. Both are ecological restoration measures and not landscape restoration.

Comments 6: Lines 20-21: The carbon decline between 2020 and 2025 deserves further contextual discussion. Were there specific public policies or climate events that explain this?

Response 6:Thank you for your insightful comments, which we fully agree with. Regarding the issue of the decline in carbon stocks in lines 20–21 of the abstract not being discussed in depth, we have provided a more in-depth discussion in Section 3.2. The original analysis was based on changes in land use types and did indeed overlook the impact of climate. We have now provided a more detailed explanation of this section, explaining the reasons for the decline in carbon stocks from 2020 to 2025 based on temperature and precipitation. The sharp decline in carbon storage (-7.69%) from 2020 to 2025 was primarily driven by changes in land use, but abnormal climate conditions during the same period also played a significant role. Compared to 2015–2020, the average annual temperature increased by 0.39°C, and the average annual precipitation decreased by 32.47 ml, jointly triggering the unique “drought-rock desertification-carbon loss” positive feedback loop in karst regions: rising temperatures exacerbate soil moisture evaporation, and reduced precipitation directly leads to drought. The shallow soil layers in karst regions have extremely poor water retention capacity, causing vegetation (especially shallow-rooted artificial forests) to suffer severe water stress, inhibiting photosynthesis and sharply reducing carbon fixation capacity. Drought leads to vegetation degradation and reduced canopy cover, making soil more susceptible to erosion. Soil erosion accelerates the exposure of bedrock, while exposed rock walls further raise local temperatures and worsen moisture conditions, forming a vicious cycle. Additionally, planted forests (carbon storage growth rate -1.31%), which rely on artificial intervention and have a single tree species, have weak resilience, while forest closure patterns (growth rate -0.97%), which rely on natural recovery and have higher biodiversity, demonstrate greater stability. This supports the conclusion that “forest closure is a more effective strategy.”

Comments 7: [Lines 243-255 (Table 1): It would be helpful to include confidence intervals or standard errors for the estimated carbon densities.]

Response 7:Thank you for your valuable feedback. We have added a specific model evaluation parameter table (Section 2.4.2) to the article. Please refer to the table below for details.

Carbon density	R²	Coefficient	Value	SE	t	p
AGC	0.642	Intercept	15.626	3.1234	5.0024	<0.01
AGC	0.642	Slope	0.838	0.066	12.7731	<0.01
BGC	0.364	Intercept	4.575	1.0131	6.9403	<0.01
BGC	0.364	Slope	0.223	0.0322	8.6266	<0.01
DMC	0.018	Intercept	1.413	0.5271	0.0406	<0.05
DMC	0.018	Slope	0.043	0.0052	0.8558	–
SOC	0.001	Intercept	149.77	144.95	15.97	<0.01
SOC	0.001	Slope	0.332	0.2849	1.1543	–

Comments 8: [Line 346 (Table 3): It is advisable to show the number of hectares involved by category, in addition to growth rates, to get a better idea of the spatial impact.]

Response 8:Thank you for your attention to detail in the article. Regarding the original Table 3, it only shows the growth rates for each year because the article explains the areas of each ecological restoration measure in Section 2.2, and since the areas of the ecological restoration measures have not changed, we did not include the areas involved in the table for land use changes from 2025 to 2030. As per your suggestion, we have added a detailed data table for land use changes, as shown in the table below.

Year	cropland	Woodland	Scrubland	Grassland	Wetland	Building land	Water
2000	43390.21	223636.52	1231.85	15745.90	432.83	1635.10	608.69
2005	42814.43	223143.89	1328.90	14964.44	1167.51	2108.96	1152.87
2010	42436.57	221818.21	1309.41	15908.67	758.41	2831.16	1621.00
2015	41109.96	219506.26	1293.72	17061.18	573.02	5020.19	2119.29
2020	41209.67	217697.82	1281.34	17867.17	483.24	5583.14	2526.70
2025	37520.50	186093.87	1244.47	35198.51	1379.95	21706.54	3488.59
2030	37159.03	183994.16	1268.44	35941.06	1184.86	22735.97	4411.37

Comments 9: [Line 374 (Table 4): Although many variables are reported, it is unclear whether a multivariate analysis or only bivariate correlations were performed. This can lead to misinterpretations due to multicollinearity.]

Response 9:Thank you for your valuable feedback. The analysis of relevance in the article was conducted using Pearson's correlation coefficient to calculate bivariate correlations, specifically the correlations between carbon storage and various factors. This is because the two main subjects, “planted forests” and “forest closure,” have a very large dataset, with hundreds of thousands of data points. Using Pearson's correlation coefficient allows for the inclusion of every data point, resulting in more accurate correlation estimates. As you pointed out, directly using the Pearson correlation coefficient for analysis may overlook the effects of interactions between factors. To address this, we considered two approaches for interaction analysis. The first was modeling, but due to the large number of influencing factors, the resulting equations and model parameters became too complex, and the modeling process was too challenging, so this method was not adopted. The second method involves using a geographic detector to analyze factor correlations and interaction effects. First, we convert the original data into point data (tens of thousands of points) and perform random sampling or stratified sampling to obtain 1,000 sample points (sampling is necessary because the geographic detector cannot process tens of thousands of data points, and 1,000 data points meet the research requirements). Alternatively, the fishing net tool was used to extract points within the fishing net for spatial connection (2,000–6,000 sample points). When performing sampling or using the fishing net tool, we reclassify most of the data into 9 categories based on natural breakpoints, while some data, such as desertification, are classified into 5 categories. Therefore, we classify these already categorized data with fewer categories into 4–5 categories. After importing the various processing results into the Geographic Detector, the correlations remain low. Even after multiple verifications and repeated operations, the correlations and interaction effects still fail to meet the research requirements. Therefore, we choose to retain the original correlation results (Pearson correlation coefficient). The Geographic Explorer results are shown as follows:

Comments 10: [Figure 5: The legend could be clearer: it is currently difficult to distinguish the carbon density curves for each restoration type.]

Response 10:Thank you for your attention to detail in the article, which greatly enhances its overall readability. We sincerely apologize for the lack of clarity in the legend of Figure 5. Indeed, the lack of distinction in the legend may have caused confusion for readers. To address this, we have revised Figure 5 by adjusting the line colors and styles, increasing the font size of the legend labels, and enhancing the PNG image resolution to 1000 DPI. This ensures that the curves representing each ecological restoration measure are now presented in a more intuitive and clear manner.

Comments 11: Line 517-520: A more complete correction of the InVEST model is recommended, or at least to discuss the implications of not adjusting SOC and DMC in the limitations section.

Response 11:Thank you for your attention to the rigor of the article. The description of the impact of limitations in Section 4.3 is somewhat vague. The soil carbon cycle has a duration ranging from 0.9 to 152 years, with an average of 24.3 years, indicating a relatively long cycle [1]. The forest closure (2006–2019) and afforestation (2009–2019) in the study did not reach the average cycle, and new trees were added annually. Forest closure primarily relied on natural recovery with supplementary artificial intervention (57.91% spot replanting), while afforestation involved comprehensive planting. The soil carbon density cycle of newly added trees was shorter than that of trees from the previous year. The study indicates that changes in soil carbon density during the implementation phase are negligible [2]. Forest litter layer decomposition and metabolism exhibit strong carbon respiration, rapid turnover, and weak carbon sequestration, with their carbon density considered a constant value [3]. The correlation between soil organic carbon and litter layer decomposition and metabolism carbon is not significant and has not been corrected. This limitation has limited impact on the study and is discussed in detail in Section 4.3 of the article.

1.Wang, J., Sun, J., Xia, J., He, N., Li, M., & Niu, S. (2018). Soil and vegetation carbon turnover times from tropical to boreal forests. Functional Ecology, 32(1), 71-82.

Reviewer 3 Report

Comments and Suggestions for Authors

The abstract is too general and repetitive. It is recommended to shorten it and clarify the main aspects: the purpose of the study (to assess the impact of forest closure on carbon storage in the karst plateau), the methods used (spatial-temporal analysis, remote sensing data, GIS), the main results (closed forests significantly increase carbon storage) and the practical significance (recommendations for land use and forest management). Excessive generalizations such as “important insights” should be avoided – they should be specified.

Introduction clarification – justification of the problem and relevance:

The introduction correctly indicates that karst regions are characterized by sensitive ecosystems and are important in the context of climate change mitigation. However, the problem should be more clearly explained: due to land degradation and unsustainable land use, carbon storage potential in these regions is often lost. Forest closure as a means of ecological restoration is often used, but its long-term impact on the carbon cycle has not been sufficiently studied, especially in terms of spatial and temporal dimensions. This would justify the importance and relevance of the study.

Justification of the choice of methods:

The methods used are sound and modern, but it would be necessary to explain more clearly why the specific time intervals and data sources were chosen - for example, whether their resolution is sufficient to assess forest dynamics in karst terrain. In addition, the accuracy of the models used (e.g. InVEST or others, if applied) should be based on literature or calibration.

Clarification of conclusions:

The conclusions could be clearer and more focused on practical solutions. For example:

Forest closure increases carbon stocks, especially during the first 10 years.

Spatial differences indicate that site characteristics (slope, soil, anthropogenic activities) have a significant impact on carbon storage.
In summary, the article is relevant and contributes to the knowledge of sustainable ecosystem management, but clearer structuring, a more concise presentation of the summary, and a stronger justification of the methods and conclusions would be beneficial.

Author Response

Comments 1:[ The abstract is too general and repetitive. It is recommended to shorten it and clarify the main aspects: the purpose of the study (to assess the impact of forest closure on carbon storage in the karst plateau), the methods used (spatial-temporal analysis, remote sensing data, GIS), the main results (closed forests significantly increase carbon storage) and the practical significance (recommendations for land use and forest management). Excessive generalizations such as “important insights” should be avoided – they should be specified.]

Response 1:[We take the issues you raised regarding the abstract section very seriously. To enhance the clarity and focus of the abstract, we will make the following revisions: First, we will clearly state the primary objectives of the study (assessing the spatiotemporal evolution of carbon storage and its driving mechanisms in karst plateau regions under different ecological restoration measures) and the current limitations of the research. Second, we will briefly introduce the main research methods employed (integrating multi-source remote sensing data and adjusting InVEST model parameters); Next, we will highlight the main research findings, namely that severe rock desertification constrains carbon storage, while afforestation can provide significantly higher long-term carbon sink benefits. Finally, emphasize the practical significance of the research and strengthen the specific expression of this part (based on the research findings, it is recommended to prioritize forest closure measures in karst regions to protect and restore forest ecosystems; simultaneously, improve local habitats and establish ecological compensation mechanisms, while strictly controlling the expansion of construction land to enhance ecosystem stability and carbon sink functions. These research findings provide a solid scientific basis for enhancing and precisely regulating the carbon sink capacity of fragile karst ecosystems, offering important guidance for formulating scientifically sound ecological protection policies, and providing specific recommendations for land use and forest management. We believe that after such revisions, the abstract will be more concise and focused, better guiding readers to understand the core content of the article.]

Comments 2: [Introduction clarification – justification of the problem and relevance:

Response 2:[We sincerely and deeply appreciate your precise control over the overall structure and content of the article, as well as the unique and profound insights you have demonstrated. Your professional perspective and valuable opinions have been of immense help and guidance to us. After careful discussion and consideration, we fully agree with and have adopted your suggestions. Currently, we have strictly followed your guidance and carefully added this insightful statement in the second paragraph of the article's introduction. This modification not only makes the article's logic more rigorous but also significantly enhances the academic value and rationality of the research, making the entire article's argumentation more powerful and persuasive.]

Comments 3: [Justification of the choice of methods:

Response 3:[Thank you for your attention to the rigor of the article; we fully agree. Based on your feedback, we have rewritten the section on data sources and presented the data in a table, including data names, resolution, processing methods, and sources. The ecological restoration measures data used in this study were obtained from the county forestry bureau's forest survey data, covering key parameters such as soil layer thickness, tree diameter at breast height, vegetation cover, tree height, and stand density; the karst desertification data were obtained from the National Karst Desertification Prevention and Control Engineering Research Center. Both the forest survey data and the karst desertification data have undergone de-identification processing and do not contain sensitive information such as topography, and are solely used for research purposes. According to research requirements, both datasets were processed using ArcGIS 10.8 to achieve a 30-meter spatial resolution, meeting the needs of the study. As per your suggestion, we have added the corresponding references after each model in the Materials and Methods section to make the article more scientifically rigorous and standardized.]

Comments 4: [Clarification of conclusions:

The conclusions could be clearer and more focused on practical solutions. For example:

Forest closure increases carbon stocks, especially during the first 10 years.]

Response 4:[Thank you for your insightful feedback, which we fully agree with. Based on your suggestions, we have revised the Conclusions section and refined the key conclusions to place greater emphasis on practical solutions. Specifically: "Ecological restoration should prioritize natural recovery methods such as close forests to enhance carbon sink stability; simultaneously, strict control over the expansion of construction land is necessary, along with the implementation of a zoned management strategy—prioritizing the protection of the high-carbon-sink ecological zones in the southwest, strengthening desertification control and human activity management in the degraded areas of the northeast, and achieving sustainable development through harmonious human-land interaction." Through this revision, the research significance of this article has been further focused on practical applications, providing valuable insights for the precise implementation of desertification control and ecological restoration projects in karst regions.]