How to Achieve the Ecological Sustainability Goal of Ecologically Fragile Areas on the Qinghai-Tibet Plateau: A Multi-Scenario Simulation of Lanzhou-Xining Urban Agglomerations

1. Summary

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.

2. Point-by-point response to Comments and Suggestions for Authors

Comments 1: This is an excellent  paper deals with what matters in present day China. The research team has sought to quantify the impacts of past policies and measure the  key indicators  that describe situation on the ground now. You identify the insidious role of excising land in name of construction land  converting some  sites to impervious and useless. Many areas extracted from farmland, grassland and forest, lie idle. The authors have shown how to adapt and develop appropriate algorithms that are robust enough to give confidence in the future projections.

Response 1: Thank you for taking the time to review our paper, and we are deeply grateful for your recognition of our work. This has been incredibly encouraging for us, and we would like to express our sincere thanks once again!

1. Summary

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions changes in the re-submitted files.

2. Point-by-point response to Comments and Suggestions for Authors

Comments 1: The title of the paper is a bit cumbersome, please simplify. 

Response 1: Thank you for your suggestions. We have simplified the title as per your advice.
The revised title is as follows: How to achieve the ecological sustainability goal of ecological fragile areas on the Qinghai-Tibet Plateau? A multi-scenario simulation of Lanzhou-Xining Urban Agglomerations (page 1).

Comments 2: There are a few problems in the chart and table. According to Table 7, it can be seen that some land use types have undergone drastic changes. Please further verify the data and give explanations.

Response 2: Thank you for your suggestions. Based on Table 7, we have noted significant fluctuations in the coverage of certain land-use categories with minimal proportions, with ice/snow coverage experiencing the most substantial variation. To address this, we re-validated our procedures and statistical results and confirmed that our operations accurately reflect the real conditions of the grid data. The data source we used is the annual China Land Cover Dataset (CLCD), which is based on 335,709 scenes of Landsat imagery processed through Google Earth Engine. This dataset provides annual land cover information for China from 1985 to 2021.

To further verify whether there are issues with the data source, we specifically examined the changes in ice/snow coverage in the Lanzhou-Xining urban agglomeration from 2000 to 2021. As illustrated in the figure, the fluctuations in ice/snow coverage exhibit a cyclical pattern, and the sudden increase in coverage in 2011 is not an isolated incident. Instead, it aligns with the average ice/snow coverage from 2005 to 2012, further validating the accuracy of our data.

We believe that the reasons for such cyclical fluctuations may include the following: First, although the yearly combination of Landsat imagery helps minimize observational bias, the satellite's 16-day revisit cycle may not precisely capture the annual maxima and the exact timing of ice/snow melt, especially in regions with significant climate and temperature variability (such as highland areas). This can lead to fluctuations in ice/snow coverage in certain years due to differences in observation timing. Second, the long-term trend of global warming, which has caused a rapid reduction in ice/snow coverage since 2018, is consistent with global climate change trends. Third, short-term climate cycles or extreme weather events in certain years may have contributed to the observed fluctuations in ice/snow coverage.

Given the large scope of this study, with ice/snow and wetland areas accounting for a minimal proportion of less than 0.0002%, and given that they are highly susceptible to seasonal and climatic factors, their fluctuations do not significantly affect the overall land cover structure of the Lanzhou-Xining urban agglomeration. We hope this explanation clears up your confusion.

Comments 3: The title information in Figure 3 is incomplete.

Response 3：Thank you for your suggestions. We have refined the title information of Figure 3 to make it clearer and more precise (line 404).

1. Summary

Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions changes in the re-submitted files.

2. Point-by-point response to Comments and Suggestions for Authors

Comments 1:  Regarding lines 123-124, what is the basic definition of Cropland preservation? Based on my literal understanding, Cropland preservation refers to the protection of agricultural land from urban expansion, industrial development, or other non-agricultural uses through policies, regulations, and measures, ensuring the land remains dedicated to agricultural production. 

Response 1: Thank you for your suggestions. In the original version of the introduction, we had provided a definition and reference for farmland protection. However, to reduce the length of the article, we later removed it. In light of your suggestion, and to avoid causing similar confusion for readers, we have reintroduced the definition and significance of farmland protection. The content is as follows: The two modes of farmland protection in China are the first focusing on protecting the quantity and maintaining a dynamic balance of the total amount of farmland, and the second emphasizing the quality of farmland in basic farmland zoning. The former proposes the "compensation system for occupying arable land" based on the revised law of the Land Management Bureau in 1998. This law requires a balance between occupying and reclaiming arable land, which largely intervenes in the selection and direction of land use for urban expansion and rural construction (lines 227 to 233).

Comments 2:  In the subsequent analysis of the prediction model, the authors do not seem to fully consider the impact of policy factors on cropland preservation. For example, the permanent basic cropland protection mentioned in line 657.

Response 2: Thank you for your valuable suggestions. In the process of setting the farmland protection scenario, based on the supplemented definition and adhering to the principles of both quantity and quality as outlined in the Land Management Law, the PLUS model can only impose constraints on the principle of farmland quantity when configuring the transition probability matrix. Referring to existing studies, we reduced the probability of farmland being converted to other land uses by 50% under the strictest farmland protection principle. Meanwhile, since the transition probability matrix already addresses the constraint on farmland quantity, we focus on the improvement of farmland quality through high-standard farmland construction in the discussion section. This is one of the reasons we chose high-standard farmland construction as a key topic in the discussion, in order to meet the policy requirements of both safeguarding farmland quantity and improving farmland quality.

Permanent basic farmland and ecological protection redlines serve as baseline requirements for regional farmland and various types of ecological land in China. These baselines are crucial for limiting land-use conversions and controlling the minimum quantities of farmland and ecological land in land-use forecasting. Unfortunately, since the LXUA spans multiple administrative regions, the vector data for permanent basic farmland and ecological protection redlines at the county level within the urban agglomeration have not been made publicly available by the government. As a result, we are unable to determine the exact area and boundaries of permanent basic farmland and ecological protection redlines within the region, which is one of the limitations of our work. Should we gain access to these data in the future, we could make more accurate land-use quantity predictions based on them. 

Comments 3: A similar issue arises with ecological protection (line 124), where it appears that the authors did not consider Ecological Red Line Delineation (line 615) in the land use prediction.

Response 3：Thank you for your valuable suggestions. In setting the ecological protection scenario, we adopted relevant spatial control requirements from the Gansu Provincial Territorial Spatial Plan and the Qinghai Provincial Territorial Spatial Plan. For ecological protection efforts, we followed the development plan of the LXUA, referring to the ecological barrier construction project in the arid mountainous areas of eastern Qinghai and the northern sand prevention belt and shelter forests. We also based the scenario on the expected growth rates of key indicators such as forest coverage, comprehensive grassland vegetation coverage, and water area retention outlined in the Qinghai Provincial Territorial Spatial Plan (2021-2035), along with the requirements for implementation in lower-level territorial spatial plans. The transfer probabilities for forest, grassland, and water areas were determined hierarchically according to these priorities. Additionally, considering the construction requirements for the quality of permanent basic farmland, we included farmland in the ecological protection scenario to safeguard the farmland ecosystem.

Permanent basic farmland and ecological protection redlines serve as baseline requirements for regional farmland and various types of ecological land in China. These baselines are crucial for limiting land-use conversions and controlling the minimum quantities of farmland and ecological land in land-use forecasting. Unfortunately, since the LXUA spans multiple administrative regions, the vector data for permanent basic farmland and ecological protection redlines at the county level within the urban agglomeration have not been made publicly available by the government. As a result, we are unable to determine the exact area and boundaries of permanent basic farmland and ecological protection redlines within the region, which is one of the limitations of our work. Should we gain access to these data in the future, we could make more accurate land-use quantity predictions based on them. 

Comments 4:  In lines 170-171, the phrase "Once the accuracy met the required standards,"—what specific standards are being referred to? Please provide references and detailed content.

Response 4: Thank you for your suggestions. This section refers to the accuracy verification results and standards, specifically the KAPPA coefficient and OA coefficient presented later in the text. In the PLUS model, we used the CLCD data from 2001 and 2011 to simulate and predict the 2021 CLCD data, and then compared the results with the actual 2021 CLCD data. Only after meeting the accuracy requirements were we able to proceed with land-use prediction experiments for future scenarios. Therefore, this section essentially serves as an early explanation of the accuracy verification process. We appreciate your suggestion, as this approach may indeed cause confusion. As a result, we have revised this section and supplemented the accuracy verification section later in the text with additional references to ensure coherence and scientific rigor in the writing (line 271).

Comments 5: In lines 197-198, "The Monte Carlo method is used when considering the neighborhood effect." Please provide an adequate explanation of the Monte Carlo method and relevant citations.

Response 5: Thank you for your suggestions. The Monte Carlo method refers to the introduction of a pseudo-random number  in the formula below to address the issue of land patch generation, allowing it to simulate the randomness and uncertainty present in real land-use processes. Following your suggestion, we have supplemented the explanation of the Monte Carlo method and added relevant citations (lines 195-197).

Comments 6:   In line 201, "?? is the threshold for generating new land use patches of type ?, defined as 0.3 in this study." Please explain why the threshold was defined as 0.3 and why the same value was used for all land uses. Since the probability of land use change is different in the Markov chain calculations (lines 207-208), it seems odd that the same value was set here.

Response 6: Thank you for your suggestions. We fully understand your concern regarding the setting of the  threshold and paste the formula here:

Where  is the operation probability of land use type  at cell  and time ;  is a random value ranging from 0 to 1;  is the threshold for generating new land use patches of type , defined as 0.3 in this study; is the diffusion coefficient of land use type;  is the growth probability of land use type  at time .

According to the definition provided in the PLUS model manual, ​​ is a user-defined patch generation threshold, with a range from 0 to 1. A higher threshold indicates a more conservative conversion strategy. Simply put, a higher ​​ value means that land-use types are less likely to undergo transformation (regardless of their respective output probabilities). It is a global strategy coefficient that interacts with a random number  to control the randomness in the process of land patch conversion or generation at a macro scale. A higher ​​ value leads the model to adopt a more conservative approach to generating new land patches, thereby controlling the overall quantity of land patch conversions. It does not interfere with the transition probability of any individual land-use type, as ​ is set separately from the land-use transition probabilities in the PLUS model interface, as shown in the figure.

In our experiment, this coefficient was determined through trial and error. That is, while keeping other simulation parameters constant, we adjusted the ​ value and observed the results to determine whether the simulated land-use changes met the expected land-use predictions. Ultimately, we set ​ to 0.3. For different land-use types, this value can remain consistent. Moreover, as shown in the figure, ​ is a global parameter in the current design of the PLUS model (likely for simplicity and operability), and separate thresholds for each land-use type cannot be set. This means that under the current model framework, we must select a unified value for all land-use types. However, this does not affect the results, as the purpose of ​—whether it is a global parameter or one adjustable for individual land-use types—is to influence randomness, ensuring that the number of generated or converted land patches meets the land-use demand during the simulation. Thus, as long as this goal is achieved, there is no need to adjust the parameter further. Adjusting parameters is merely a means to an end, not the end itself. 

The Markov chain, on the other hand, aims to predict future land-use demand by adjusting the probability matrix to calculate the land-use demand for different scenarios. The final output is entered at the location shown in the figure. The Markov chain determines the future transition probabilities of various land types, while ​ primarily controls the number of randomly generated land patches. The two functions operate independently of one another. Thus, ​ does not affect land-use change probabilities, and vice versa.

While we have not encountered a situation in this experiment or in previous studies where adjusting ​ values for each land type was necessary to meet the land-use demands, we do not rule out the possibility. However, such a case would be highly complex and would significantly complicate the parameter adjustment process, making it challenging to manage.

Lastly, we realized that placing the parameter results obtained through trial and error directly into the formula could cause confusion, and the term " defined as" was also inappropriate. To avoid such misunderstandings, and in reference to similar studies, we have removed this statement from the methodology. Thank you for your suggestion (line 201).

Comments 7: Additionally, part of the content in Formula 5 seems to contain errors. If there are n types of land use, the result should be n.

Response 7: Thank you for your suggestions. In this formula, we are representing an  transition matrix ​, where  and  represent different land-use types. The expression  means that the sum of the transition probabilities for each land-use type  equals 1, which is a constraint in the Markov chain model. It is not calculating the number of land-use types. Therefore, the result should be 1, not . We have modified the expression of the matrix formula to make it more standardized. The transition matrix ​ is represented as:

We hope these explanations help to clear up any confusion (line 207).

Comments 8: In lines 243-246, the authors only use Kappa and OA indicators to evaluate the model. It is recommended to add the Figure of Merit (FoM) indicator to enhance the credibility of the results.

Response 8: Thank you for your suggestions. Following your advice, I used the PLUS model to calculate the Figure of Merit (FoM) and have included it in the manuscript. The FoM value for this experiment is 0.7156 (line 271).

Comments 9: Regarding Table 7 in line 372, does "Area" refer to land changing from other uses to this land type, or from this land type to other uses, or both? Please clarify.

Response 9: Thank you for your suggestions. The areas mentioned here refer to the actual areas of various land-use types in different years, as well as the rates of change in land-use areas between different years. This does not involve land transfers. To avoid such misunderstandings, we have optimized the table names. We hope this explanation clears up your confusion (line 384).

Comments 10: In line 404, "Ecologically, the BS scenario leans towards natural restoration." I am unable to understand from the analysis how the authors arrived at this conclusion.

Response 10: Thank you for your suggestions. The wording here was indeed inaccurate. What we intended to convey is that the changes in various types of ecological land under the BS scenario continue the land-use change patterns observed over the past twenty years. We have revised this statement accordingly (lines 419-420).

Comments 11: Lines 416-418, "However, this increase in cropland is spatially manifested as a delay in urban construction and more aggressive encroachment on grassland." And lines 500-501, "the disorderly expansion of cropland." These statements suggest that the model's prediction results are not consistent with normal land protection policies, but rather point toward extensive agricultural land development.

Response 11: Thank you for your suggestions. As mentioned earlier, I have supplemented the definition of farmland protection in this paper, which includes the principles of both quantity and quality protection. The CP scenario simulated using the PLUS model emphasizes the binding objective of the first principle—improving farmland quantity by reducing farmland loss—without controlling the conversion into farmland. This results in the spatial phenomenon of farmland encroaching on grassland and expanding in a disordered manner. As you pointed out, this trend focuses on agricultural land development, but under the constraint of China's farmland protection policies, such phenomena do indeed exist. This is because the boundary between agricultural land development and farmland protection is not clearly defined in China. For example, the farmland requisition-compensation balance policy is regarded by the government as a major measure for farmland protection, but the policy itself involves developing new farmland. This is especially true in western regions where per-unit yield is generally low, leading to more aggressive expansion efforts to increase farmland area (i.e., newly compensated farmland exceeding the amount of occupied farmland) and, thereby, overall yield.

Moreover, in recent years, high-standard farmland construction (which focuses on improving farmland quality) has included large-scale land consolidation measures, such as “contiguous plots, land leveling,” and preventing farmland fragmentation, which also involve reclaiming new farmland and expanding existing farmland. In the context of China's emphasis on food security and the significant loss of farmland due to urbanization over the past decades, the net increase in farmland area is often seen as a key target and political achievement for local governments in their farmland protection efforts. Therefore, under the dynamics of controlling farmland quantity, the extreme emphasis on food security, and local governments’ pursuit of farmland net growth as part of their performance evaluation, agricultural land development is inevitable in China’s farmland protection efforts.

Your understanding of farmland protection as a policy, law, and measure to safeguard agricultural land from urban expansion, industrial development, or other non-agricultural uses, ensuring that the land remains for agricultural production, aligns well with China’s current permanent basic farmland system. This system serves as a baseline control mechanism to ensure that land within designated permanent basic farmland areas is not converted to other uses. However, in China, permanent basic farmland accounts for approximately 83.1% of the total farmland area, and as a baseline control measure (ensuring basic food security), it may not be aggressive enough for broader farmland protection, especially in scenarios where the primary goal is farmland protection. Achieving net farmland growth may better align with the performance expectations of local governments and the policy direction that the Chinese government has upheld in recent years. We hope this explanation clears up your confusion.

Comments 12: There is a spelling error in formula 9: "expoential" should be "exponential."

Response 12: Thank you for your suggestions. We have corrected this spelling error (line 313).

Comments 13: In line 317, "??? is habitat degradation" should be placed below formula (10), i.e., under line 307.

Response 13: Thank you for your suggestions. We have made the necessary correction (line 320).

Comments 14:  The North arrow is missing from Figure 3 in line 388.

Response 14: Thank you for your suggestions. We have made the necessary correction (line 404).

Comments 15: The units are missing from the legends in Figures 6 and 8.

Response 15: Thank you for your suggestions. I have added unit information to Figures 6 and 8(line 483 and 545).