The Effects of Terrain Factors and Cultural Landscapes on Plateau Forest Distribution in Yushu Tibetan Autonomous Prefecture, China

Round 1

Reviewer 1 Report

The paper presented by Naixin Cui, Huiting Zou, Moshi Zhang & Luo Guo for publication in Land titled ‘The effect of terrain factors and culture landscapes on plateau forest distribution in Tibet Plateau: a case of Yushu Tibetan Prefecture, China’ seeks to investigate the role of topography (elevation and slope) and landscape use on the area of forest habitats in the Tibetan Plateau and presents some interesting results. The authors find a strong elevation and slope control on forest area and a substantial increase in forest area in close proximity to temples in the study area.  Whilst at heart this paper has some very interesting results, I have detailed several major points of concern in the analysis and presentation of the manuscript along with a many minor points.

Major issues:

1. Forest classification and change detection methods:

I am very pleased to see that the authors use remote sensing data to quantify forest change across the prefecture and link these changes to both environmental and anthropogenic factors. However, there is vital information missing from the manuscript that is essential to fully understand the forest classification and change data that forms the primary dataset investigated in this manuscript.

1. It is unclear what data is being used to create the forest classification maps and how the authors have arrived at a forest change dataset. Two datasets are mentioned in the text, historical forest maps from RESDC, and unspecified ‘images’. The authors state that images were classified using visual interpretation and supervised classification methods with ‘optimization’ using field data. At present the information provided in section 2.2. is insufficient. As a minimum I would like to see the following specific details:
   1. The classification method used should be named, including the number of training and testing points, any cross-validation parameters, or method specific parameters that are necessary to enable another researcher to carry out the same procedure.
   2. Specific information about the ‘Images’ used in the classification. If these are the forest landscape maps previously described please explain why a classification was required and what information the forest landscape maps contain to give the reader sufficient context.
   3. Classification statistics are missing, and a confusion matrix detailing class specific user and producer accuracies should be presented alongside the overall accuracy that has been detailed. The confusion matrix is essential as there is substantial scope for error in classification that occurs between forest classes which would undermine the forest classification maps used for the subsequent analysis.
   4. The information given in sections 2.3.3 and 2.3.4 suggest that a single land-use change map has been used in the authors analysis but there are no details provided to explain how this land-use change map has been created or what is contained within the change map. Please clarify if a change-map was used and if so, provide the necessary details for the readers to follow your methods more easily.  
2. The forest definitions presented by the authors lack the necessary detail to describe how they are defined. At present the authors define four classes, however the ‘shrub forest’ is defined as land with ‘canopy density greater than 30% and less than 40%’ which is completely contained within the ‘forested land’ class which is defined as ‘canopy density greater than 30%’. If there is another parameter (e.g. height or species) that is used to separate these classes, then please provide details otherwise the forest classes will need to be redefined so that they are discrete

2. Regression analysis:

I am unclear what data have been used to carry out the regression analysis. In section 2.3.1 the authors state that five altitude and slope classes were used, and the forest class areas were extracted from these elevation or slope bands. If my interpretation is correct, the authors have only used 5 data points in each forest class for each time step to carry out their regression analysis which equates to a single data point per each elevation/slope class for each forest type which is not a sufficiently large enough sample size for environmental regression analysis. As such I have the following recommendations depending on if my interpretation is correct:

1. If I am incorrect, please provide additional details as to which data is used in the regression and where it has originated from. Also present the raw data in the figures and not and average point that intersects with the fit line. The figures should also include a confidence interval.
2. I my interpretation is correct, I suggest that the authors carry reconsider their choice of analysis or increase their sample size by taking a large number of random sample across the gradient of slope and elevation within the study area and calculate the area for each virtual plot. I would also urge the authors to take careful note of the papers authored by Pontus Olofsson and colleagues that specifically address challenges obtaining reliable area estimates from image classifications and the importance of sampling for estimating habitat areas. The authors of the papers listed below describe methods for area estimation based on visual interpretation of land cover change maps and remote sensing data. Crucially these methods will provide a measure of uncertainty which appears to be missing from the current analysis.

       I would recommend the following papers as a starting point if the authors         choose to follow this path:

• Olofsson et al. 2016. Time series analysis of satellite data reveals continuous deforestation of New England since the 1980s. Environmental Research Letters. 11, 064002.
• Olofsson et al. 2014. Good practices for estimating area and assessing accuracy of land change. Remote Sensing of Environment. 148, 42-57.
• Olofsson et al. 2013. Making better use of accuracy data in land change studies: Estimating accuracy and area and quantifying uncertainty using stratified estimation. Remote Sensing of Environment. 129, 122-131.

3. Language and Style:

I found this paper very difficult to follow because the authors frequently use terms that are not defined, overly technical or don’t have any specific meaning to the field alongside very complex sentence structures. In combination the complex language and structure have the effect of reducing readability and interpretability and, therefore, may have led to some of the confusion I have surrounding your methods. In addition, the figures are of poor quality and the legends are lacking detailed information and must be improved so that the figures can be interpreted without the main body of text. I have provided some examples of specific details on language and presentation in the minor comments below. However, the authors should make every effort to substantially reduce the complexity of their sentence structure, define all technical terms that they use and improve the overall flow of their manuscript.  

Minor points:

Abstract:

Line 19: The acronyms RS and GIS are used without being introduced.

Line 23: What laws are referred to when the authors write ‘The four types of forest landscapes show certain laws with the change of terrain, and these laws are significantly related’

Line 24: What are the laws significantly related too?

Line 26: Which forest type conforms to which distribution function?

Line 28: In what context is ‘horizontal distribution’ used? Nowhere in the manuscript is horizontal defined. The term has many potential meanings here including longitude or the land perpendicular to the slope. Please avoid ambiguous terms and define your use of terms that are not commonly used in the literature. I would advise you to avoid the term horizontal if possible. 

Line 29: What does ‘more positive influence’ mean in this context?

Line 30: In what way is ‘village influence on forests limited’? If it is the case that there is less forest habitat in village landscapes then it could be argued that the village has a significant detrimental influence on forest areas.

Line 33: Please state which which terrain factors you have investigated.

Line 33: Again the use of vertical distribution is ambiguous here. Please rethink your use of the terms Horizontal and Vertical throughout the manuscript.

Introduction:

Line 59: Avoid using author first names for in text citation.

Line 80: Please define what you mean by planar and triaxial in this context. There is many examples in the literature where of forest change research is predicted based on multiple environmental drivers which seems to be what you are trying to infer with your use of triaxial.

Line 85: Define what the ‘pattern problem’ is.

Line 86-87: Please explain where the gap you refer to lies and why the method you propose is suitable for filling it. Many of the methods you refer to earlier in the paragraph can make use of both spatial and temporal information and so it is not clear where there is a gap and how a simpler regression approach would fill it.

Lines 89-94: Your aims are all contained within a single sentence with multiple clauses. In this current form they are not easily understood and should be simplified to aid readability.

Methods:

Line 104: The term perennial does not seem to fit with the context of the sentence. Inter-annual may be more appropriate here.

Line 105: Please use scientific notation of units of temperature.

Line 118: Have the previous field investigations referred to been published?

Line 130: What is the accuracy of the DEM?

Line 133: Is the DEM sufficiently accurate to discriminate between the slope values into such narrow bands. When converting a slope from percent to degrees I would expect your range of slope values (written as 0-27%) to range from 0 to 15°. As such your first class is just 0-1° while the second class is ~1-3°.  I would also have expected to see greater slope values in the mountainous areas surrounding the plateau and so could you please provide some descriptive statistics to complete the readers understanding of the study environment and where the forest habitat exists.

Line 152: Here you only consider three forest classes but you previously described four classes. Please be explain why you drop the fourth class from some analyses and not others and be consistent throughout your manuscript.

Results:

Figure 2: This figure is of low quality and it is difficult to read the axes and legends. The figure also appears to be missing the x-axis label and the elevation seems to go from high to low but should progress logically from low to high in the figure. The plot also appears to be missing the raw data with only average values presented.

Figure 3: Please avoid 3-dimensional plots when point data are used as they are difficult to interpret. The figures could easily be displayed in two dimensions with year represented as different shapes.

Line 224: the term ‘all other forest’ is confusing here. You have a class called other forest but here and elsewhere you blur the line between your defined forest classes and talking about the forest and a single entity. Please revisit the naming of your forest classes to ensure they are clear and be explicit when you refer to a forest class and when you are discussing forest habitat generally. For example you could put 'the other forest class’ in this instance.

Line 229: You conclude that the shrub forest class is the most widespread class, however, because this class overlaps in definition with the forested land class it is not possible to make this conclusion without a clearer separation of the forest classes and assessment of the classification accuracy confusion matrix which may highlight if there are any other issues in class separability that would impact your area estimates.

Line 249: Please present the effect sizes from your regression analysis in the tables and main body of text. Without the effect size the information given doesn't support the statement that there is a negative relationship.

Figure 4: The variable (radius) used in this figure are not circular and so spider diagrams are inappropriate here. By changing the layout so that Forest area is on the y-axes and radius is on the x-axis it will be much easier to understand the data being presented. If there is little or no change year to year, I would also suggest you only present the data from a single year so that it is easier to interpret.

Figure 5: This figure is very difficult to interpret due to the poor quality and small size of the individual elements.  If there is little or no change year to year, I would suggest you only present the data from a single year o that it is easier to interpret.

Table 4: Please provide confidence intervals for your area estimates and also the area estimates for the villages.

Discussion:

The discussion in it’s current form is problematic because it primarily repeats the results without any substantial interpretation or relation to the existing literature, there is only a single reference in the discussion. As such a more detailed discussion is required that ground the results presented in the context of the wider literature on controls of forest distribution and the influence of environmental and cultural controls.

Please discuss additional topographic controls that you have not investigated here. Aspect, curvature, sheltering (e.g. topographic index position) alongside many other topographic features have all been shown to influence forest distribution globally and so should at least be discussed if they are not included in the analysis. The discussion of the cultural landscapes on forest distribution is better but still requires more references and better comparison to existing research.

Avoid the use of the term ‘causal relationship’ throughout your manuscript. As an example, the current use of the term from line 311 is incorrect. The manuscript reads ‘The distribution of the three forest landscapes on the altitude gradient has the same law, and in the regression analysis, forested land and sparse land are fitted with Polynomial Function, and shrub land is fitted with Logarithmic Function, which further illustrates the causal relationship between altitude factors and forest distribution.’ However, the conclusion that elevation causes the observed pattern in forest distribution is not valid since elevation itself does not determine the forest distribution. Instead elevation a proxy for other environmental factors, namely temperature, which are co-linear with elevation and are the major control on forest distribution globally.  

With best wishes.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Dear authors, 

I have gone through the manuscript and found authors did very nice research and coming under the scope of the journal. But some of the major corrections are being needed before considered for publication:  

Minor comments: 

Kindly rewrite the Abstract in concise way (not more than 250 words) 

Line no: 22: an altitude of 5055 m and 6300 m (it should be 5055 masl and 6300 masl, kindly check and correct thoroughly in manuscript)

Line no. 31: kilometers should be written like Km (kindly correct it thoroughly, written in uniform way)

Line no. 48-52: sentence need support with references (see suggested one)

Line no. 61-67: sentence need support with references (see suggested one)

Line no: 85, CLUE-S Model....needed to be explain a bit 

Line no: 105: temperature 2 degrees centigrade (kindly use symbol like oC)

Line no: 133-134: parts, including 0_2%, 2_6%, 6_15%, 15_25%, 25_27%, and 3568_4333 m, 4333_4624 m, 4624_4828 m, 4828_5055 m, 5055_6300 m...............Kindly coreect it like 0-2%.........3568-4333 masl.

Line no. 132 and 144 is not same (ArcMap 10.4.1 or ArcMap 10.4 ??)

Line no:152 : we use SPSS for common correlation analysis....kindly write the SPSS tool no

Equation numbers has to be given to each equation (e.g., Eq.1, 2, 3,... )

Increase the resolution of figure 2

Discussion section need support of past literature, kindly read it once and support the confidential statements 

Conclusion need to be written in technical way instead of generalized 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

This study focused on the effects of slope/elevation and culture landscapes on plateau forest in the Tibetan plateau. The second part, i.e., the effects of temple and village on forest distribution is interesting. However, the first past is nothing new. In my opinion, this paper can be improved substantially. More comments are as follows.

--Title. This study is not a method study and it is not right using "a case of ...". The right title is "The effects of Terrain factors and culture landscapes on Plateau forest distribution in Yushu Tibetan Autonomous Prefecture, China.

--Abstract. Too long, and the logic can be improved.

--Introduction. Line 45, "Describing" should be "describing". Line 61-63, previous studies just focus on one factor, and "there is less research (researching) about the effect of both natural and cultural factors on forests." But in this study, the authors just conducted the effects of slope/elevation on forest, and then the effects of temple/village on forest, not the mutual influence of multiple factors on forest. So this study is just a combination (paper A: slope/elevation, paper B: temple/village, and this paper: A+B) and nothing is new. What the readers expected is the mutual influence of multiple factors on forest, and the contribution rate of each factor, not just A+B.

--Study area. Confusing. Yushu Tibetan Autonomous Prefecture, or Yushu Prefecture, which one? They are totally different. Line 91, 98, Yushu Prefecture, but line 123, Yushu Tibetan Autonomous Prefecture. Please check and revise.

--Figure 1. Yushu Tibetan Autonomous Prefecture, or Yushu Prefecture? The spatial scope is Yushu Tibetan Autonomous Prefecture. But the legend is Yushu Prefecture. Only DEM is shown, which is not informative. The distribution of temple is welcomed.

--Line 133-134, Why is altitude and slope so divided? What is the basis?

--Line 140-141. the Qinghai geographic information center. Reference(s) is/are needed.

--Methods. sections 2.3.1 and 2.3.2 focused on terrain factors (which is nothing new), and 2.3.3-2.3.5 focused on temple/village. In my opinion, terrain related methods and results are not necessary.

--Figure 2. Very fuzzy. The font size of the axis is too small. Line 219, 5 periods should be 6 periods.

--Figure 3. (a) is a duplicate of Figure 2 and should be deleted. and both (a) and (b) are very difficulty to read and to find the change tendency.

--Figure 5. It is really difficulty to find the differences among the six subfigures. Additionally, the study area is missing in the figure caption.

--Discussion. What's the advantages of this study compared with previous studies, and the limitations?

Author Response

Please see the attachment.

Round 2

Reviewer 1 Report

The revisions provided by Naixin Cui, Huiting Zou, Moshi Zhang & Luo Guo are welcomed. However, unless I have received an incomplete version of both the manuscript and response letter, the authors have omitted certain points from my review that referred to key methodological concerns. In addition, the authors have said language experts have made changes but these changes do not appear to be present in the version of the manuscript I have received.

The revisions made to the methods remove reference to remote sensing data and now suggest that all analyses are based on the Landcover maps provided from the RESDC, comparing the forest cover of each year individually rather than as a single change map. Additionally, the authors have also said in their response letter that the results from the land-use change analysis were not included in the paper. Despite this the authors still refer to their use of remote sensing data and land-use change data in the abstract introduction and methods, in context of the methods and results presented in this paper. For example, the authors write ‘In a given series of radius distances rq, the Pc(rq) spatial relationship function of a specific land use change can be expressed as’  yet no methods are presented that detail how the land use change has been derived and the results appear to report the forest cover in each year separately and not land use change data. Please clarify the text throughout the manuscript readers are clear what data has or has not been used. If land-use change data have been used please present the methods.

The purpose of the regression analysis and the data used in the analyses has not been addressed in the manuscript with the authors also omitting some of my original comments from their response. The manuscript suggests that the regression is primarily implemented as a descriptive piece of analysis to identify the mathematical function that cumulative forest distribution takes. There is no problem with this, but it should be clearly stated in the aims because the regression carried out only has five data points which are not sufficient for hypothesis testing. This is particularly problematic as the authors still refer to a ‘causal’ relationship between terrain and forest distribution in both the introduction and methods section, although they have removed the term from the discussion as previously requested.   

Minor outstanding points:

Line 156 states ‘The data were preprocessed using’ Please detail which pre-processing steps were applied to both the Land use and elevation data.

The forest classes have been updated as requested but are not completely defined in the text. On line 171 Closed forest land is still defined as all forest above 30% canopy cover while the response letter states it was 30-40%. Closed has also been misspelt throughout the manuscript where the new land cover label has been copied.

The authors response to my previous question about the definition of slope is fine but the reference should also be presented in the manuscript and not just to myself. As you use a Chinese national standard for the slope classification it would be helpful to describe the standard in the text to avoid future confusion for the international audience of this journal. Within the text please include the description of the slope classes including the degree of inclination that is contained within each percentage class.

There are still instances where the authors refer to using either three or four forest classes without making clear in the manuscript which forest class has been dropped.

Figure 2 is still very low-quality resolution and elevation goes from high to low on the x axis and ought to be reversed so that it is presented from low to high, left to right. The forest class labels within the figure also no longer match the forest class names used in the text and need to be updated.

The authors removed Figure 3 in response to a previous comment. I should clarify that I think Figure 3 should be left in but should be presented in a different way. Figure 3 is potentially more useful than figure 2 but at present it is difficult to interpret due to the 3-d presentation.

Figure 4 is still problematic; it is of low-quality resolution and the labels do not match the text and so must be updated. I appreciate that the authors wish to present all years and not just one, but I would still prefer the spider diagrams to be updated with classical x-y plots so that it is more intuitive for the reader to understand the change being presented.

Section 4.1 line 345. While I appreciate that this text was added in response to a previous comment of mine, I am unsure why the authors have introduced carbon storage at this point. In particular, the distinction between the distribution of forest habitat and the distribution of forest carbon storage has not been made clear. In my opinion, it would be better to avoid the simultaneous discussion the distribution of forest habitat and the distribution of forest carbon as there are many biologically sound reasons why they may differ. Instead focus on the habitat distribution which follows the accepted understanding of global forest distribution.  

Author Response

Please see the attachment.

Author Response File: Author Response.doc

Reviewer 3 Report

I cannot find the author’s response to my comments. The attachment file is empty. So please do not reply my comments by uploading attachment file. To be honest, this paper do not improved much. My concerns are still exist.

--Abstract. Still very long.

--Introduction. Do not improved much.

--Line 133-134, Why is altitude and slope so divided? What is the basis?

--Figure 3. (a) is a duplicate of Figure 2 and should be deleted. and both (a) and (b) are very difficulty to read and to find the change tendency.

--Figure 5. It is really difficulty to find the differences among the six subfigures. Additionally, the study area is missing in the figure caption.

Author Response

亲爱的评论员，

Thank you very much for your time and comment on our manuscript. 

Below we document our responses to your comments. Our responses start with "Response:" and are highlighted in red. The line numbers used here refer to the ones in the Word document (with Track Changes turned to "All Markup"). All line numbers are based on the revised manuscript unless stated otherwise (e.g., "in the previous submission").

The comments put forward by you include the following aspects, which we answered respectively:

1.Abstract. Still very long.

Response: Thanks for your comment on manuscript. We have briefed the abstract section. Please see the abstract section of the revised manuscript. (Lines 11-33).

2.Introduction. Do not improved much.

Response: Thanks for your comment on manuscript. We have improved the introduction and invited professional language editors to revise the full text to reduce the complexity of the sentence in revised manuscript. ( Please see lines 68-127 in revised manuscript ).

3.Line 133-134, Why is altitude and slope so divided? What is the basis?

Response: Thanks for your comment on manuscript.The slope gradient classification we used is the Chinese standard classification, which was divided into five levels: 0–2°, 2–6°, 6–15°, 15–25°, 25–27°. In Yushu Tibetan Autonomous Prefecture, the upper limit of slope is 27°, so the fifth slope gradient is 25-27°.We have divided the altitude gradients based on previous research. For a detailed explanation of the method, please refer to Cui,N et al (2018) and TD/T 1055-2019 (2019).

"Cui,N.; Luo, G.; Du, S. Analysis of Spatial-Temporal Variation of Grassland Landscape Pattern Based on Terrain Factors in Qinghai Yushu Tibetan Autonomous Prefecture, China. The 26th International Conference on Geoinformatics. 2018."

"Ministry of Land and Resources of the People's Republic of China. TD/T 1055-2019, Technical regulation of the third nationwide land survey. http://gi.mnr.gov.cn/201901/t20190129_2391915.html"

4.Figure 3. (a) is a duplicate of Figure 2 and should be deleted. and both (a) and (b) are very difficulty to read and to find the change tendency.

Response: Thanks for your comment on manuscript. We have revised Figure 3 to be more intuitive. (Please see line 273 in revised manuscript)

5.Figure 5. It is really difficulty to find the differences among the six subfigures. Additionally, the study area is missing in the figure caption.

Response: Thanks for your comment on manuscript. Figure 5 shows the changes in the forest around temples, but compared to the area of Yushu Tibetan Autonomous Prefecture, the area of the buffer zone is small, so it is indeed difficult to see the changes visually. The data in Table 3 can see that the forest changes every year, but the distribution of the forest cannot be directly seen, so we want to visualize the data in Table 3 so that readers can see the distribution. We have revised Figure 5, improved its clarity, and added the study area to the legend.