If They Come, Where will We Build It? Land-Use Implications of Two Forest Conservation Policies in the Deep Creek Watershed

Round  1

Reviewer 1 Report

In general the article is well written and the research achieved interesting results. The readibility of the article can be improved (suggestions made in detailed comments) and the authors need to justify better why they used the CLUE-S model with its limitations. I think that some of the most important observations need to receive more attention: lines 276-279; 309-316; 462-464.

Comments for author File: Comments.docx

Author Response

See attached file.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments for forests-529397

This paper used the CLUE-S model to simulate land use change under some policy scenarios in a small-scale watershed. My overall suggestion is that the paper should be improved by justifying the methodology adopted for the research. This by no means challenges the model itself, but rather requires more details on data-processing, table preparing with probabilities of transition, and their associated uncertainties. Please see the specific comments below, which primarily aim to encourage the authors to clarify some confusions. 

In the section of Calibration, please provide reasons on why aggregated land use types into the five categories, particularly clarify the difference between the three confusing classes: 1) livestock farm; 2) forest and range land; 3) pasture and forage land. For example, I would think of combining rangeland and pasture in terms of their function, but it appears that the authors categorized them based on their properties (natural vs anthropogenic). Also, livestock farms may overlap with rangeland or pasture in some cases. 

I am also wondering why using the 500 by 500 spatial scale. Clearly, the statement that “it leads to about 1000 cells” is not a reason. The scale issue has been discussed by many studies, as inappropriate scale selection can be misleading in analysis involving spatial issues. For instance, land use patterns with a high spatial resolution may reflect additional information on consequences due to policy (or policy combinations) that land use with a coarse resolution would not show. On the other hand, the coarse resolution can provide general patterns that are more useful for policy design while the fine resolution can be associated with too much noise. Please discuss more. 

Please also add more details on how to derive the transition probability impact table based on the logistic models. First, consider adding a section of data source after the section of study site, and providing full descriptions on the data such as source, unit, spatial resolution and/or any issues (e.g., missing values). In the method, clarify how these data in various format were resampled or rescaled to fit in the 500m by 500m grids. There are plenty missing or unclear information. For example, for distance to urban center, which location(s) is/are defined as the center? Should distance to paved road be precisely distance to the nearest paved road? What is spatial association, and why is it excluded for forest and range land and included in others? Why are aspects so important that they need to be included to predict the probability of transition? Is there any theoretical justification on selecting these variables for transitions among the land use types? These are just a few questions of many. Without the relevant details, the research would not be reproducible or validated. These details also pertain to uncertainties (or limitations, caveats, etc.) of analysis, which should also be thoroughly discussed in Discussion Section. 

Section 3.1 fits in the section of method. This section describes how to design the 4 scenarios that are the combinations of the two conservation policies. Similarly, the first paragraph in Section 3.3 should be placed in methods too. In addition, it is unclear about the two analyses (4 policy combinations versus ALR) and the scenario design for ALR. It sounds to me that ALR is implemented within a specific boundary, but the model excluded this bounded area (i.e., ALR area), and then did the simulation for the regions (very small part as the authors stated) only outside the ALR. This would lead to the results on how much pressure the outside small regions can bear with land use change. But the authors discussed the “pressure on lands within ALR”, which is confusing. 

Why delineate a boundary based on the elevations below 550m? Is this for the purpose of defining the valley area, as mentioned in the figure caption? This may be better introduced in the main text, either in study sites or method. Then, why it is necessary to compare the valley bottom with the ALR, while these two appear to cover nearly the same region of the landscape. 

Author Response

See attached.

Author Response File: Author Response.docx

Round  2

Reviewer 2 Report

I very appreciate the revision made by the authors, which to a substantial extent clarified most of the confusing points. For example, the justification of using the CLUE model over others (e.g., agent-based model) in this study has been discussed and articulated in methods; land use categorization has been clarified for the model inputs; the reason of scale of interest for the landscape has been explained; the detailed procedures of deriving the predictors for the transition probabilities have been stated (e.g., measured distance); while the limitation of the CLUE mode was acknowledged, its application in land use projection and implications for conservation policies has been strengthened.

In addition, the authors address issues that are less closely tied to the findings of the land use modeling effort in this study, such as the terms of complex interactions, coupled human-natural systems (line 506), feedbacks, payment for ecosystem service (line 474) and agent-based analysis [e.g., 1,2]. The “feedbacks” seemed at most only implicitly examined by the current investigation in this paper, which would be more possibly simulated by other models (e.g., agent-based model) and thus may need to be less focused in discussion. That being said, for example, may consider deleting the last sentence in Abstract, as it reads that the advocate of the agent-based approach undermines the rationale of using the CLUE model. Some helpful references are suggested to consolidate the discussion on such human-environment topics involving individual decision-making [3]. In line 133, may consider citing more references [e.g., 4] to justify the application of CLUE for modeling land use dynamics. In line 417, some recent references may also be useful to address the complex relationships between human and the landscape processes [5-7]. Finally, PES refers to payment for ecosystem services as stated in line 474 and associated with forest conservation policies, making them so-called incentive-based policy (line 476). I hope these examples are useful to address the comment on “discussion more on what has not been achieved than what has been achieved”.
Overall, minor suggestion is recommended. I look forward to reading the paper in a published format when online.

References
[1] Walsh, S. J., & Mena, C. F. Interactions of social, terrestrial, and marine sub-systems in the Galapagos Islands, Ecuador. Proceedings of the National Academy of Sciences, 2016, 113(51), 14536-14543.
[2] Tian, Q., Holland, J. H., & Brown, D. G. Social and economic impacts of subsidy policies on rural development in the Poyang Lake Region, China: insights from an agent-based model. Agricultural Systems, 2016, 148, 12-27.

[3] Zhang, Q., Song, C., & Chen, X. Effects of China’s payment for ecosystem services programs on cropland abandonment: A case study in Tiantangzhai Township, Anhui, China. Land use policy, 2018, 73, 239-248.

Author Response

See attached file.

Author Response File: Author Response.docx