Quantifying the Spatial Association between Land Use Change and Ecosystem Services Value: A Case Study in Xi’an, China

Round 1

Reviewer 1 Report

The paper presents a case study on associations between land use/cover dynamics (changing speed) and ecosystem service (value) variations. Despite the revision, my general sense is that misinterpretations and confusions persist in this paper. Please see below some of the specific points focusing on the methodology.

The selection of 1 by 1 km grid is not convincing. It is simply based on the arbitrary sayings, and thus needs to be appended with the detailed outputs of the comparison among different sizes. In addition, what does it mean by non-grid process? Does it mean the original Landsat spatial resolution or treating the whole study area as the unit? If analyzing the whole study area, how to make the comparison, as the grid-level outputs are not aggregated? Also, how to handle the mismatch of resolutions of Landsat and the redefined grid size of 1 km, and how that leads to uncertainties? These details are rather vague or missing.

The calculation of the comprehensive changing rate is confusing. Equation 3 obviously points to the outputs corresponding to the land cover types, meaning each land use type should have one value of the comprehensive index. As far as I am trying to understand, the component of the denominator is the sum of all land use types, which is the total area of the study area (assuming filled up with all land use types), a constant. Accordingly, this comprehensive index measures, for a specific land type (e.g., farmland), the annual changing percentage of only conversions from this given type to others. This is contradictory to what has been introduced in the previous text, raising much confusion for readers.

The processes of constructing the GWR is not clear. First, the paper seems never to mention how to decompose the ESV into the grid-scale. Based on the model description in the ESV section, the outputs are not related to grids that could lead to results of spatial heterogeneity. Second, how to sample across the space to estimate the GWR (and OLS) model, and what is the sample size, etc.? It is really required to provide as much detail (e.g., additional steps and associated assumptions) as possible to make the research design clear so one can further evaluate the reliability of the results.

Miscellaneous incorrectness should be avoided. For example, in equation 2, LU_a and LU_b should be specific to one land use type, so need to have subscript of i. The unit for unit area yield should not contain economic units such as yuan (Line 184).

Please note that these are only a few examples of many careless presentations; by “careless” I sincerely mean it, which is a very critical problem. It is the author(s)’ responsibility to ensure the precision and rigorousness of the whole paper, and contribute to the journal with a high quality manuscript. Be clear and precise about the details, be explicit to address the assumptions, and be thoughtful when discussing the uncertainties. Do not hide the “weaknesses” or cite others without proper thinking about the research design itself. Please take the revision seriously, not just relying on the editors/reviewers’ voluntary help.

English writing:

Improving language should not only simply correct those typos or grammatical errors (which still exist in the text notwithstanding), but more importantly also focus on the logic flow. Below are specific examples regarding the abstract, aiming to encourage critical thinking and thorough checking for the full text.

In abstract, study area that is to be divided is not introduced at all; not until later of result interpretation can readers guess that it is in Xi’an in Line 24.

Following the major finding “the distribution of ESV showed spatial heterogeneity”, one should expect solid evidence to support the heterogeneity, i.e. how ESV vary across space; yet woodland providing half of ESV is not relevant.

Continuingly, what does it mean by the negative association between LULC and ESV? Does it mean “higher degree of land use change with decreased ESV” or “directional change such as from farmland to urban (assumed a positive change of this direction) with decreased ESV”. In the former case, the fallacy exists in that some change can be associated with increased ESV, such as farmland to forests; in the latter case, the confusion is rooted in the unclear definitions in methodological design.

It should be the total ESV loss (n.), not lost (adj.).

Author Response

Dear reviewer :

We would like to thank for you valuable feedback that we have used to improve the quality of our manuscript.(ID:sustainability-766700), We had checked the manuscript and revised it based on your opinions. The data processing, research methods, and conclusions have been carefully modified.All grammatical issues have been revised directly and marked in red in the revised draft. The reviewer comments are listed in black text and specific concerns have been numbered. Our responses are given in red text. Point by point response to the reviewers’ comments are listed in response letter.

Thanks again for the precious comments from you.

Best regards,
   Sincerely yours

 Yajing shao

Author Response File: Author Response.pdf

Reviewer 2 Report

This study purports to quantify the spatio-temporal pattern of ecosystem services value (ESV) in Xi'an over the period 2000-2018. Overall, the study is highly descriptive, with a huge number of large tables, several multi-plot figures, and long paragraphs devoted to their description. But the value of this descriptive detail is unclear and no attempt is made to synthesize this detail into new hypotheses or policy implications.

I have serious concerns about the way ESV is calculated in this study. The authors state: "the economic value of one ecosystem service equivalent is the average economic value of food production in one hectare of farmland ecosystem..." The implication is that ESV for the variety of functions listed in Table 1 (including climate regulation, hydrological regulation, biodiversity, and "aesthetic landscape") can all be found as a scalar multiple of the value of exactly one function: food production. Yet a very similar study in China found that "the circulation of materials and energy flows in ecosystems, which decide the diversity of the ecosystem service, are extremely complex" (Song et al. 2015). Why should the value of food production determine the value of such unrelated and diverse ecosystem services?

This also calls into question the relevance of calculating ESV at all once land-use change in quantified. Since ESV is assigned as a fixed multiple (Equation 5) of food production (Equation 4), aggregate changes in ESV can be reduced to changes in land use. Because the method used assumes that ESV of any ecosystem service in any land is some fixed multiple of the value of potential food production, the summary statistics of ESV that are reported simply reflect the change in balance of different land uses over time.

I strongly urge the authors to fundamentally reconsider the methods used. I do not believe the prior work by Xie et al. is relevant or useful in this study. The authors should re-design the calculation of ecosystem service value so that it accurately reflects the different values of the diverse functions they have listed. Only then can they accurately calculate the spatio-temporal variation in ESV. If the authors insist that this methodology is adequate, they must completely re-frame the study and justify this decision: why should we value a particular non-farm land use and its variety of ecosystem functions solely according to its ability to produce food?

Lines 37-38: These are "types of" services (or "classes" of services), not services in themselves.

Lines 55-64: This is the first and last time we hear about the Benefits Transfer Methods (BTM); it is not mentioned in the Methods section of this paper. I am left wondering if BTM is used in this paper or if the methods used are in anyway related to BTM. Or does this paragraph serve simply as Background on the valuation of ecosystem services in general? If it's just background, it should probably go in a separate Background section after the Introduction. If it's NOT Background and is directly related to the novel aims, methods, or findings of this paper, please make a direct connection between BTM or other prior methods and the innovations in this paper.

Lines 56-58: Assuming that BTM is relevant to the methods used in this paper, given that ecosystem services include "cultural services" (Line 38), is BTM really valid for extrapolating ecosystem services value (ESV)? If the reference assessment is conducted in one cultural or political context, can it be used to provide accurate ESV in a different cultural or political context? For example, one population may value a suburban housing density differently than another; or they may value rivers, forests, or farmland differently. Are the approaches to non-market valuation for cultural functions different than for other (e.g., regulation) functions that may be more tangible or measurable? De Groot et al. (2002, Ecological Economics) identifies four broad valuation methods: direct market valuation, indirect market valuation, contingent valuation, and group valuation. Perhaps the authors could (likely in a Background section) provide some justification as to their approach (if and how it is related to BTM). From this Introduction, we get a sense of *how* ESV might be assessed or transferred to a new context, but not why this is justified or how ESV is assigned in the first place.

Lines 71-82: Could you provide some concrete examples as to how spatial dependence or heterogeneity complicate ESV assessment? For spatial dependence, I might imagine that the ESV of a forest is higher if it is closer to a dense human settlement. For heterogeneity, I might imagine that a given spatial unit delivers higher ESV if it contains multiple regulating functions as opposed to only one. For spillover effects, I can imagine you might want to estimate the marginal ESV or ESV *net of* the ESV provided by adjacent lands. Perhaps it is for these ()and other) reasons that you are asserting that attention to spatial dependence and heterogeneity are important; but concrete examples, particularly if they are supported by prior literature or case studies, would be more compelling.

Lines 89-92: What does it mean to "quantify dynamic land use change at a grid scale"? Are you interested in the *rate of change* in land-use? Please be more explicit. Similarly, you aim "to explore the spatial association between LULC change and ESV at a grid scale;" what spatial association in particular? How will it be explored? Is this simply a matter of cross-tabulating LULC and ESV changes?

Lines 117-122: It seems you may be using a supervised classification method here but it is not described. I gather that you "[manually] interpret[ed]" land use from the imagery. Please specify how land-use was assigned to each pixel in each year in your study area. What were the covariates used in a supervised classification? Was it spectral data (Landsat TM/ OLI bands) only or did you use additional data such as elevation to inform the classification? How were the classification training samples chosen?

Lines 130-132: "The attribute data after grid processing can not only ensure data integrity, but also fully represent the spatial consistency characteristics of attributes." I have no idea what this means.

Lines 134-138: On what basis did you decide that "the analysis results at a 1 km grid scale were generally consistent with results obtained by non-grid processing"? I think it would be valuable to compare the effect of scale (e.g., 1-km, 2-km, or 3-km grid cells) on your results if space and time permit. But if picking an optimal scale for gridding is the objective, maybe it would be more appropriate to use variogram analysis to determine the scale of spatial autocorrelation and grid accordingly? Please also explain *why* gridding is needed; this wasn't clear from Lines 130-132, but I would believe that one reason would be to reduce the computational complexity of processing the data at pixel (30-m) scale.

Lines 166-168: It is now apparent that BTM and the prior assessment of ESV by Xie et al. is very relevant to this paper. As Xie et al.'s assessment was for the same study area, BTM seems appropriate. However, I think it would be good to succinctly state here *how* ESV was assigned in the Xie et al. studies. For example: "The ESV of each potential land use and service type was determined by Xie et al. through a combination of hedonic modelling from land transaction records, case studies of extractive indu and focus group interviews." How Xie et al. arrived at the assessed ESV by land-use class is central to your study since you provide summary estimates of the total value or change in value across your study period.

Line 172: What is "urban public mining" land? Is this two different land types? Should there be a comma here somewhere?

Line 170-173: I would argue that "woodland," "grassland," and "water bodies" are land-cover types, not land-use types. Are the "woodlands" and "grasslands" actively managed for, e.g., timber production or ranching? Perhaps it would be better to refer to these, throughout the paper, as "land-use/ land-cover" (LULC) types. There are referred to as "ecosystems" here, which adds to the confusion. Why does "farmland" include "dry land" (what is "dry land" in this context)?

Line 173: If "unused land" refers to "bare land" I would strongly recommend it be called "unused bare land" to avoid confusion. Certainly at least some of the "woodlands" and "grasslands" are also "unused" (i.e., not actively managed for resource extraction, grazing, etc.). Seeing as "unused land" area is tiny compared to any other land use, you might consider ignoring it in this study.

Lines 173-174: I am very skeptical that "developed land" provides zero ESV. Recent studies in environmental justice, urban ecology, urban planning, and public health studies do not make this assumption (e.g., Walker et al. 2017; Ashton 2015; Schwarz et al. 2015; Haase et al. 2014; Wolch et al. 2014). Bolund & Hunhammar (1999) delineate multiple urban ecosystem services in an economics journal. Moreover, a very relevant study of ESV in China also does not assign zero value to "built-up" land (Song et al. 2015). Since developed land is assigned zero ESV, I think it is important for you to clarify your definition of ESV. You define ecosystem services by listing different types (Line 38) but, clearly, not all of these types inform your definition of ESV, as developed lands must provide some non-zero cultural functions. Perhaps it is simply a matter of whether developed lands are considered ecosystems? It's not necessary to get too far into the weeds here, but it would help if you could specify exactly why developed lands have zero ESV--perhaps it is because you assume they are completely hardscaped/ completely impervious surface, which would provide no biogeochemical regulation benefits. This would seem to imply ecosystem services are faciltated only by "natural" systems, yet farmland (not natural) has non-zero ESV for many functions/ services. As a side note, spatial scale influences this decision because a 1-km grid cell certainly has some sub-grid land-use and land-cover heterogeneity. In short, I believe a much stronger justification for assigning zero ESV to developed land is required. It is certainly possible that developed land provides *less ESV* for a given function compared to other land types, but I don't see any evidence that value is near zero. If the goal is to compare more "productive" (in ESV terms) land types with developed land, one possible change to the study that avoids assigning zero ESV to developed land would be to specify the ESV of other land types relative to that of developed land. In such a case, one *might* find that the relative ESV of all other land types is a positive value for any function, i.e., all other land types provide higher ESV for a given function than developed land types.

Lines 176-193: "The economic value of one ecosystem service equivalent is the
average economic value of food production in one hectare of farmland ecosystem, and equal to 1/7 of the market value of the crop yield per unit area." Why should this be so? Equation 4 relates to only one type of land (farmland) and one type of ecosystem service (food production). You have listed several different ecosystem functions in Table 1 that have nothing to do with food production, yet the implicit assumption is that they have value. Is it simply not practical or possible to calculate the ESV of functions other than food production? Song et al. (2015) write: "The circulation of materials and energy flows in ecosystems, which decide the diversity of the ecosystem service, are extremely complex;" diverse ecosystem function require diverse approaches to valuation. Equation 5 seems to suggest that the ESV of any given land type and ecosystem function is simply a scalar multiple of the ESV for food production in farmlands. The use of "equivalent coefficient data" from Xie et al. (2003) requires further justification. Functions like "soil conservation" might rightly be compared to the *export* of organic material that is food production. But why should functions like "hydrological regulation" or "climate regulation" be valued the same way? Nowak et al. (2006), for instance, assigned a value to the air pollution removal function of urban tree canopy by calculating "the estimated cost of pollution to society that is not accounted for in the market price of the goods or services that produced that pollution." There is some logic to calculating ESV as a flux (e.g., organic material or nutrients, water) but that flux should be *relevant* to the ESV in question. Zhang et al. (2012), for instance, calculate the ESV of green space for stormwater reduction based on the unit prices of reservoir and reclaimed water. Song et al. (2015) use primary production for the calculation of the value of "provision of organic material" but use "the average cost of reservoir construction" to value water conservation.

Lines 196-197: In the rest of the text, there is no distiction between ESV by land-use (Equation 6) and ESV by function (Equation 7); the text just reports "ESV". Is it necessary, then, to distinguish between these two functions? I can only assume that Equation 6 is the only relevant equation going forward.

Lines 204-206: I think it is more accurate to say that GWR recognizes the spatial heterogeneity in *relationships* between variables and the spatial non-stationarity in *relationships* (not "geographic features"). The two sentences that start this paragraph would seem to be saying the same thing, so one of them could probably be deleted.

Lines 209-210: "The GWR model is a spatial extension of the least squares method, but it is slightly inadequate in the diagnosis of the model." Could you be more explicit? Are you referring to model diagnostics like (global) goodness-of-fit or t-statistics for model terms?

Line 256: Table caption should be changed to "Comprehensive land use dynamic degree in..." to distinguish this from "single land use dynamic degree."

Line 266 and Figure 2: Figure legends should be made larger as they are difficult to read.

Lines 289-292 and Figure 3: Figure legends should be made larger as they are difficult to read.

Line 312 and Table 5: The "Total" of "Proportions" should be 100, not 1, as they are given in percentage terms.

Lines 331 to 333: This paragraph is in conflict with an earlier statement. Here, you write: "this study took the CLUDD in 2000−2005, 2005−2010, 2010−2015 and 2015−2018 as dependent variable, and the change of ESV in each period as the independent variables." But on lines 207-208 you write: "The dependent variables in this study were changes in ESV, while the independent variables were changes in CLUDD."

Line 335 and Table 6: While the dependent and independent variables need to be clarified (see above), this table would seem to suggest that the CLUDD in each period forms the set of independent variables. What hypothesis is being tested by regressing change in ESV on CLUDD in each period? Am I to interpret the coefficient values as the amount of reduction in ESV attributable to each period? Or are there 4 different models? It would seem to be the latter ("for every unit increase of CLUDD, the ESV was reduced by 749,688.71 units") but what is the relevance of this statistic, then? This should be better motivated by explicitly stating research questions and hypotheses earlier, around Lines 89-92.

Line 343: I think "non-stationarity" rather than "non-equilibrium" is what is meant here.

Lines 357-362: Instead of reporting both the R^2 values ("0.95, 0.66, ...") and writing "the GWR model explained 95%, 66%, ..." you could remove one of these sentences; they convey the same information.

Lines 416-418: You write: "Our results indicated that woodland contributed the most to ESV in Xi'an, accounting for nearly 50% of the total ESV in the study area from 2000 to 2018. Therefore, woodland degradation contributed most to ESV reduction." Yet, according to Table 5, the ESV of woodlands decreased by only 5.07 million yuan compared to farmlands, which decreased by 183.34 million yuan. Woodland ESV as a percentage of total ESV is basically the same in 2000 and 2018 (50.18% and 50.57%, respectively). Obviously, it is loss of farmland that contributes the most to ESV reduction over 2000-2018. Looking at Table 3, it's also clear that farmland experienced the greatest conversion, in land area terms, to other land uses. This begs the question as to what value there is in quantifying ESV in this way. Since ESV is assigned as a fixed multiple (Equation 5) of food production (Equation 4), aggregate changes in ESV can be reduced to changes in land use; we're forced to accept that unwritten coefficients from Xie et al. are correct because they alone determine differences in ESV between different land use types.

Lines 428-429: "From 2000 to 2018, the association between LULC change and ESV showed a significant negative relationship." A major reason for this is the conversion of other land uses to developed land, which was assigned an ESV of zero. We could have reached this same conclusion simply by looking at a land-use change map for 2000-2018, knowing that "ESV" is defined as zero in newly developed lands. Why are the extensive tables, maps, and regression models needed?

Lines 454-456: "When quantifying the ESV, a revision was made according to the regional characteristics, which can only reflect regional average states." Can you elaborate on this? What revision was made?

Lines 471-473: I would avoid using the term "SLUDD" in the Conclusion as it makes it harder for a casual reader to assess and understand your work. The phrase "rate of land-use change" could easily be used instead.

Lines 484-485: "A significant negative association was documented between LULC change and ESV change." Again, I question the salience of this point given the way ESV is calculated.

Lines 485-487: "The GWR results demonstrated strong spatial variations in time and location of the effect of LULC change on ESV." The purpose of GWR in this study requires much more motivation. While the dependent and independent variables are unclear, it seems that change in ESV was regressed on the "CLUDD" or the "rate of variation for all land-use types within a certain period." CLUDD is in percent units(?) so the coefficients would indicate the amount of ESV change for a 1% change in land-use area. If this coefficient varies across space, isn't it entirely due to the different mix of land-use changes across space? The use of GWR is an interesting tool for visualizing the effect of these different land-use mixes, but I think it's value in testing or generating hypotheses here is overstated.

Lines 481-482: You write: "The total ESV of Xi'an City decreased by 0.3 billion yuan from 2000 to 2018, representing a decrease of -0.81%, mainly because of the ESV decrease of farmland." This contradicts your earlier statement on Line 418 that "woodland degradation contributed most to ESV reduction." Of course, the tables show that it is farmland that contributed the most.

Minor typographical/ grammatical edits:

Line 37: "including" should probably be "include"

Line 149: What units are "hm^2"? Square hectometers? Is this a typographical error and it should square kilometers (km^2)? Hectometers are an uncommon choice, so you might want to write it out (unabbreviated) the first time.

Line 160: For clarity, I think the word "respectively" would help here.

Line 192: "VCif" should be "VC_{if}" with "if" in subscript notaton. Also, a sentence shouldn't be started with an equation.

Line 415: "Woodland play" should probably be "woodlands play"

References cited

Ashton, J. 2015. Plants and green spaces provide more than just aesthetic benefits. Perspectives in Public Health 135 (4):178–9.

Bolund, P., and S. Hunhammar. 1999. Ecosystem services in urban areas. Ecological Economics 29 (2):293–301.

Haase, D., A. Haase, and D. Rink. 2014. Conceptualizing the nexus between urban shrinkage and ecosystem services. Landscape and Urban Planning 132 (August):159–169.

Nowak, D. J., D. E. Crane, and J. C. Stevens. 2006. Air pollution removal by urban trees and shrubs in the United States. Urban Forestry & Urban Greening 4 (3–4):115–123.

Walker, C. M., K. Colton Flynn, G. A. Ovando-Montejo, E. A. Ellis, and A. E. Frazier. 2017. Does demolition improve biodiversity? Linking urban green space and socioeconomic characteristics to avian richness in a shrinking city. Urban Ecosystems 20 (6):1191–1202.

Schwarz, K., M. Fragkias, C. G. Boone, W. Zhou, M. McHale, J. M. Grove, J. O’Neil-Dunne, J. P. McFadden, G. L. Buckley, D. Childers, L. Ogden, S. Pincetl, D. Pataki, A. Whitmer, and M. L. Cadenasso. 2015. Trees grow on money: Urban tree canopy cover and environmental justice. Plos One 10 (4).

Song, W., X. Deng, Y. Yuan, Z. Wang, and Z. Li. 2015. Impacts of land-use change on valued ecosystem service in rapidly urbanized North China Plain. Ecological Modelling 318:245–253.

Wolch, J. R., J. Byrne, and J. P. Newell. 2014. Urban green space, public health, and environmental justice: The challenge of making cities “just green enough.” Landscape and Urban Planning 125:234–244.

Zhang, B., G. Xie, C. Zhang, and J. Zhang. 2012. The economic benefits of rainwater-runoff reduction by urban green spaces: A case study in Beijing, China. Journal of Environmental Management 100:65–71.

Author Response

Dear reviewer :

We would like to thank for you valuable feedback that we have used to improve the quality of our manuscript.(ID:sustainability-766700), We had checked the manuscript and revised it based on your opinions. The data processing, research methods, and conclusions have been carefully modified.All grammatical issues have been revised directly and marked in red in the revised draft. The reviewer comments are listed in black text and specific concerns have been numbered. Our responses are given in red text. Point by point response to the reviewers’ comments are listed in response letter.

Thanks again for the precious comments from you.

Best regards,
   Sincerely yours

 Yajing shao

Author Response File: Author Response.pdf

Reviewer 3 Report

Title: Quantifying the spatial association between land use change and ecosystem services value: A case study in Xi’an, China.

Manuscript ID: sustainability-766700 (second revision)

The original manuscript was deeply amended with both rearrangements and integrations. The readability greatly improved, several doubts were clarified, and critical features or limitations overcome. Now, I’m satisfied with this work and just a few details remain to be adjusted.

38-41: Please, consider the following rearrangement:

“Land use and land cover (LULC) change plays an important role in the formation and functional allocation of ecosystem services. Indeed, it is the most direct manifestation of the interaction between human activities and ecosystems functioning, and one of the main forces driving changes in the value of ecosystem services.

43: Please, consider the following suggestion: “which will affect the supply of ecosystem services”

46-47: Please, consider the following suggestion: “In this context, quantifying the relationship between LULC change and ESV”

48-49: Please, consider the following suggestion: “can provide a reference for the regional optimization of the land use pattern and sustainable development”.

53: Please, consider the following suggestion: “land planning”.

83: Please, delete the following sentence, it is repetitive (see line 88-89) “This study attempts to reveal the spatial association between LULC change and ESV”. 88-89: “Our study explored the spatial relationship between LULC change and ESV”.

90-92: “quantify dynamic land use change at a grid scale during 2000 to 2018; and to explore the spatial association between LULC change and ESV at a grid scale in Xi’an city”. The Authors are still using the following expression: “grid scale”. As written in the first review of the manuscript, the expression “grid-scale” does not make sense. I can understand that a grid was superimposed to the land-use map to extract the data, but the grid itself can express different scale values concerning the chosen mesh dimensions. You frequently use this incorrect expression all over the manuscript; I warmly suggest avoiding this use, by deleting or finding a different solution.

114: “generated”; I would prefer “retrieved”.

121: I still have doubts about the term “developed” land and I think it should be clarified. Furthermore, the term “unused” land seems strange to me and should be clarified if it is still natural land (i.e. without any men interference) or “abandoned” land, instead (i.e. land left over by men).

130-140: the explanation about the choice of a grid having a 1x1 km mesh is still unclear. I suggest to avoid any kind of explanation and directly state that the analysis was performed with that kind of mesh. Please, delete the following text: “The attribute data after grid processing can not only ensure data integrity, but also fully represent the spatial consistency characteristics of attributes [35]. During grid processing, the research area was divided into 1×1 km, 2×2 km, and 3×3 km grid units using the grid scale method in a GIS (ArcGIS 10.5) and the Create Fishnet tool. Because the size of the smallest unit may influence the research results, we compared the characteristics and differences in the spatial and temporal pattern of ESV at different grid scales, and found that the analysis results at a 1 km grid scale were generally consistent with results obtained by non-grid processing, and could express the characteristics of ESV distribution in the study area”.

158: Equation (1), and line161, Equation (2). The unit of both SLUDD and CLUDD is [% y-1], i.e. per cent per year. In the two formula appears 100%. This is not correct, simply remove %.

167-168: Please, clarify this sentence: “According to the level of ecosystem development and social economy, the ecosystem services in Xi’an city were divided into four level one and 11 level two categories”. What is the meaning? You also use the expression” ecosystem development”. What do you mean?

201: Table 1. I suggest aligning the numbers to the right, as well as inserting the comma to express the thousands.

205: “This model can better characterize the spatial non-stationarity of geographic features”. Please, modify the expression “non-stationarity” with a more appropriate word.

2019: “random variable”. Better to say “the error random term”.

Figures 2, 3, 4 and 5. Five land cover categories were selected in all the figures. The question is: According to what kind of statistical distribution these categories were defined? What is the quantile corresponding to each land category? Is this quantile always the same in every figures?

The figure captions should be significantly improved and made more understandable. Furthermore, please specify the units of the mapped variables in the captions.

331-333: “this study took the CLUDD in 2000−2005, 2005−2010, 2010−2015 and 2015−2018 as dependent variable, and the change of ESV in each period as the independent variables”. Well, to be honest, it is the opposite! Exactly as you wrote on lines 207-208: “The dependent variables in this study were changes in ESV, while the independent variables were changes in CLUDD”.

339-342. Please remove the following sentence, it is clear from Table 5 and there is no use to specify every figure. “The results of the OLS model indicated that for every unit increase of CLUDD, the ESV was reduced by 749,688.71 units. From 2010 to 2015, the CLUDD has the least impact on the ESV. For every unit increase of CLUDD, the ESV decreased by 330,298.04 units”.

343: “To account for the spatial heterogeneity and non-equilibrium of land use dynamic changes”. Please, avoid using the term “non-equilibrium”. You are measuring the LULC changes but you do not know what the equilibrium is.  

357-362: There is no reason to repeat every figure reported in the table. A shorter sentence is more appropriate. Please delete the following sentence: “In the four periods, the goodness of fit of the GWR model reached 0.95, 0.66, 0.90, 0.97, 357 and the adjusted goodness of fit was 0.95, 0.66, 0.89, and 0.97, while the goodness of fit of the OLS model was only 0.83, 0.33, 0.45 and 0.85. This means that in 2000–2005, 2005–2010, 2010–2015, and 2015–2018, the GWR model explained 95%, 66%, 90% and 97% of ESV changes, respectively, with an average value of 86.75%. However, the OLS model only explained 83%, 33%, 45% and 85% of ESV 361 changes at most, with an average value of 61.50%”.

374, 383: please, avoid the term “speed” and prefer “rate”. See also line 151.

373, 379, 382, 384, 385, 390, 392, 394, 432, 435, 450. The Authors repeatedly used the term “correlation”, but a regression analysis was performed. The two statistical analyses are different and should not confused: correlation means covariation of two variables, while regression means functional dependence of a variable (y) on another variable (x). Therefore, I warmly suggest avoiding using the term “correlation” and using the right statistical expression.

409: “The spatial characteristics of the CLUDD reflected that human interference and economic development was not in balance in Xi’an city”. Please, avoid the term “in balance” for the same reason the term “non-equilibrium” was criticized.

Author Response

Dear reviewer :

We would like to thank for you valuable feedback that we have used to improve the quality of our manuscript.(ID:sustainability-766700), We had checked the manuscript and revised it based on your opinions. The data processing, research methods, and conclusions have been carefully modified.All grammatical issues have been revised directly and marked in red in the revised draft. The reviewer comments are listed in black text and specific concerns have been numbered. Our responses are given in red text. Point by point response to the reviewers’ comments are listed in response letter.

Thanks again for the precious comments from you.

Best regards,
   Sincerely yours

 Yajing shao

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

I appreciate the attempt of clarification. Here quotes one of the responses, “In the OLS model and the GWR model, the sample size is 10,522, that is, the grid is used as the research unit for regression analysis.” It turns out that the whole population of grids, instead of randomly selected samples, was used for the regression models. It is no problem running ols or gwr with population data, but there is no meaning to stand error, t-statistics and p-values, which are shown in Table 6 as an example. The estimated parameter is the real relationship for the full population of grids, no matter the p-value is “statistically significant” or “statistically insignificant” (assuming statistical reference is forced to be used here for population data since it is based only on sample). These statistical issues are rooted in the very beginning of methodological design and discourage the further review of result interpretation. It might be “unrealistic” for the authors to start it over or take out meaningless results, but it is definitely impossible to let this issue slip off. Thus, it is urged to clearly put this quoted description in methods section 2.3.4, and meanwhile add more discussion to address this issue in section 4.3. I hope the “minor suggestion” will not discourage the authors to make a diligent revision at this round and craft a well-designed paper in future research or on-going work. As one of the many potential readers, I would read through the paper if it is online.

Author Response

Dear reviewer :

Thank you very much for your valuable feedback that we have used to improve the quality of our manuscript. We had carefully checked the manuscript and revised it based on your opinions. All changes in the article are highlighted in the marked manuscript. Point by point response to the reviewers’ comments are listed in response letter. The reviewer comments are listed in black text and our responses are given in red text.

Thanks again for the precious comments from you.

Best regards,
   Sincerely yours

 Yajing shao

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors have made some serious improvements to the paper. Most notably, I am very glad that the different LULC types have been rationalized and that developed lands are not arbitrarily assigned zero value. However, in my opinion, the authors have failed to engage with many of my comments in the last round of review. In particular:

1. The authors have not stated why their method is superior to that of recent and highly relevant papers (e.g., Zhang et al. 2012; Song et al. 2015), some of which were conducted in similar demographic and economic settings, but which used more diverse sources of value for transfer; e.g., Song et al. used "the average cost of reservoir construction" to value water conservation. While these other papers use more diverse, direct, and relevant accounting of non-market values, the authors here use only food production. One has to wonder why.
2. What, again, is the motivation for investigating "spatial association" in LULC change and ESV? The authors write frequently about "spatial association" but once LULC in each year is accounted for, the model of ESV is completely determined by the table of values per unit area. Any "spatial association" allegedly in ESV is actually a spatial association in LULC change: the Moran's I results indicate that LULC changes by type or by overall degree are spatially clustered. If GWR shows the coefficients vary across space, assuming the result is not simply due to a change in the domain of the independent variables (not checked for here), we could arrive at the same conclusion--that the period of intense LULC change varies across space--by looking at the maps (Figure 3).
3. The authors' chosen method for calculating ESV still needs to be justified and described in full. Throughout the paper, they describe as "results" conclusions that are completely determined by their initial valuation of different LULC types (Table 1), which is not described in full but derived from publications that are generally not available to readers. For example: "Our results indicated that woodland had the greatest impact on the ESV of Xi'an City" (Lines 420-421). This is because, as the authors state earlier, "This was mainly because the woodland in the study area covered a large area and had a high ESV coefficient." If the authors want to claim this as a "result" of their study, they should at least describe the initial valuation in full. I made recommendations on how to do this in the line edits below.
4. In the Response to Reviewers, the author's write: "Based on the reviewer's opinion, we added a hypothesis in the background section." It is highly unusual to add hypotheses to a paper after it has been written. The hypothesis has allegedly been added to the "background section" but there is no background section in the revised manuscript and the quoted line numbers (lines 98-99) do not seem to be correct (these lines are a description of the study area). Concrete examples as to how spatial dependence or heterogeneity complicate ESV assessment were asked for and changes were made to lines 82-86, but these changes are limited to a single statement ("LULC change is affected by factors such as nature, the economy, and policy and has strong spatial heterogeneity and spatial spillover") that only tacitly repeats the importance of spatial heterogeneity and spillovers without providing any clarification. Neither economic regulation nor policy implementation are mentioned in detailed hypotheses or formally modeled, so this statement would seem to amount to a hand-waving argument. I urge the authors to revisit the Introduction to make it clear what the research questions and hypotheses are.

Song, W., X. Deng, Y. Yuan, Z. Wang, and Z. Li. 2015. Impacts of land-use change on valued ecosystem service in rapidly urbanized North China Plain. Ecological Modelling 318:245–253.

Zhang, B., G. Xie, C. Zhang, and J. Zhang. 2012. The economic benefits of rainwater-runoff reduction by urban green spaces: A case study in Beijing, China. Journal of Environmental Management 100:65–71.

It's clear the authors believe that the prior literature that uses this method (Xie et al. 2003; Zhang et al. 2015; Peng et al. 2016; Kang et al. 2019) is sufficient justification for their own use. I disagree; of these four references, only Kang et al. (2019) is truly available to readers and Kang et al. does not justify or describe the method in sufficient detail. The burden is on the authors to fully describe the principle of transferring "the economic value of...food production" to all other ecosystem services for the first time in the mainstream literature.

I think the authors must make the following changes to justify their methodological approach:

• The authors should make it explicitly clear in the Introduction and Conclusion that this is a particular, normative framing of their research: all ecosystem service values are just some multiple of the value of food production.
• The authors should include Table 1 from the Response to Reviewers (the table of "ecological equivalent coefficients") in the main text.
• The authors should describe in detail how the "ecological equivalent coefficients" were determined; this should not require readers to read the papers by Xie et al. (2003), Zhang et al. (2015), or Peng et al. (2016) as these papers are not accessible (see my comment on Lines 164-181, below).
• The authors should fully justify their decision--after Kang et al. (2019) who also did not justify the decision--to set "The per-unit value of ecosystem services was set as 1/7 of the total economic value of food production service from cultivated land" (Kang et al. 2019). What is this physical meaning or economic significance of 1/7?
• The authors should make it clear that any "spatial association" between ESV and LULC change is due to the *type* of LULC change and the spatial variation in dominant LULC types and modes of LULC change.

Line-by-line comments:

Lines 62-63: Because the methods used are based on transferring values relative to one ecosystem function (food production) in one type of land (croplands), I really must insist that this sentence be amended to read: "...which laid a foundation for quantifying the ESV in China relative to food production in croplands." Or "agricultural lands."

Lines 85-86: Again, this sentence really should be revised to read: "...between LULC change and ESV relative to food production in Xi'an City."

Lines 164-181: If the authors insist on continuing with this method for transferring ESV, much more explication is needed. The benefits transfer method (BTM) is a well-established but very general approach. I take issue with the specific choice, independent of BTM, to tie valuation of diverse ecosystem functions in diverse LULC types to a single function in a single land type (crop yield in croplands). The estimation of crop yield may be very well identified but the authors write (in the Response): "The natural food production is generally proposed to be 1/7 of the actual food production." What is "natural food production" as opposed to "actual food production"? Are we talking about the amount of biomass that can be recovered for human consumption--that this is 1/7th of the overall biomass that could be harvested? The references for this figure (Xie et al. 2003; Zhang et al. 2015; Peng et al. 2016) appear to be in regional journals... I could not access the full text for any of them on Google Scholar or Mendeley, nor could I find the journals home pages; on Mendeley's page for the Xie et al. (2003) paper, the link that supposedly provides the full text is linked to a completely different paper and the abstract in Mendeley is for a third paper altogether. The Kang et al. (2019) also cites Xie et al. (2008) but this paper is also not available on Scopus (not even the abstract) [1] nor on Google Scholar, which has only one link to the supposed source material [2], which is hidden behind what I assume is a paywall (the site is exclusively in Chinese). Costanza et al. (1997), cited earlier, does not provide a justification of "equivalent coefficients;" rather, their assessment was based on "either: (1) the sum of consumer and producer surplus; or (2) the net rent (or producer surplus); or (3) price times quantity as a proxy for the economic value of the service..." and this study is not described in the same terms. Suffice to say... I don't think it's likely that readers of this paper will be familiar with the methods used here nor with Xie et al.'s methods; Kang et al. also neglected to specify or justify any real details of the method, writing simply "the per-unit value of ecosystem services was calculated with the economic value of the food production service from cultivated land per hectare." I think the principle of transferring "the economic value of...food production" to all other ecosystem services needs to be justified and the method fully described in this paper, not left to an obscure prior literature: How are the "ecological equivalent coefficients" determined (w_{if} in Equation 5)? In the Response to Reviewers, the authors noted that the coefficients were developed after consultation with "more than 700 ecologists;" this is one of the details that should be included in the manuscript, not just the Response. The table(s) of ecological equivalent coefficients must be included in the manuscript (they were included in the Response to Reviewers but should be in the manuscript). Why is the "the per-unit value of ecosystem services...set as 1/7 of the total economic value of food production service from cultivated land" (Kang et al. 2019); why not 1/6 or 1/9?

[1] https://www.scopus.com/record/display.uri?eid=2-s2.0-72149120605&origin=inward&txGid=75d43d85f489f24e47893020c0882ffa
[2] http://en.cnki.com.cn/Article_en/CJFDTotal-ZRZX200805019.htm

Lines 194-198: I think the reason the authors are using NDVI here is because different lands have different levels of productivity, perhaps based on soil characteristics or even climate, which is not otherwise reflected in the valuation. This is good, but it should be motivated in the Introduction, around Lines 90-91. The authors write "NDVI can reflect the conditions of regional climate and natural resources" but I think describing this in terms of productivity is more appropriate, since the value transfer method is (regrettably) based on food production.

Line 208 and Table 1: What is "evolutionary environment" and how does it differ from biodiversity? Costanza et al. (2019) had "pollination," "refugia," and "genetic resources" as three different services. I think "biodiversity" in Table 1 refers to genetric resources, possibly refugia, but I'm not sure how "evolutionary environment" maps on to this classification.

Line 208 and Table 1: Please add column totals to this table, since the total by LULC type is what is used in Equation 5.

Lines 359-362: I think you should make it clear that you are describing the average R^2 value across periods; "...the GWR model explained 85% of ESV changes. However, the OLS model only explained 55.75% of ESV changes." should be "...the GWR model explained 85% of ESV change, on average. However, the OLS model only explained 56% of ESV change, on average."

Line 403-404: In what year was the "Grain for Green" project implemented or over what years was the program in effect? Was this program available throughout the study area? The authors note a positive relationship between ESV change and LULC change "in the southern area of Zhouzhi, Huxian, Chang'an, and Lantian in 2000–2005" but it's not clear what might be the cause, except perhaps due to "a faster transition between woodland and farmland" (Lines 437-439). The authors should make more explicit any potential connections to economic policies as drivers of LULC change.

Lines 456-458: I would like to see acknowledged the limitation of value transfer based on a single ecosystem service. For instance, you might change these lines to read: "We calculated the ESV of diverse ecosystem functions and LULC types by weighting them relative to crop production. Based on this method, the ESV relative to crop production can be assessed easily and quickly, though it may not accurately capture willingness-to-pay for ecosystem services or consumer or produce surpluses in value."

Lines 476-478: The authors write: "the ESV was modified by the [NDVI] to show the spatial heterogeneity of the ESV in different grids." I think this is a misrepresentation of what NDVI is purported to show. Including NDVI provides information on spatial heterogeneity in biomass/ productivity across the study area, which *may* be related to differences in potential ESV. It's important to be clear about this because: (1) The choice of NDVI needs to be motivated, i.e., *why* NDVI was chosen over other spatially varying quantities; and (2) Because the transfer of a single ESV (food production) to diverse ecosystem functions and LULC types is a critical (and tenuous) assumption.

Lines 495-496: The authors write: "LULC changes have significant negative effects on ESV. At the same time, the relationship between LULC changes and ESV changes with time and space." I think it is misleading to characterize "LULC change" as some monolithic quantity of interest--it is always the *type* of LULC change that matters. In this study, the only reason that "LULC change" may increase total ESV in one place or decrease total ESV in another (prior to the inclusion of NDVI data, though it's unclear how much of an affect this has) is because of the difference in the balance of LULC types and direction of change across the study area. The conclusion that "LULC change" has some complex relationship with ESV over time and space--given that ESV per unit area is fixed by LULC type--is akin to the conclusion "changes in the size of meals and changes in weight vary with diet;" a diet that consists mainly of simple carbohydrates (sugars, starch) may increase weight but a diet that consists of high-fiber, low-calorie complex carbohydrates (vegetables, whole grains) may decrease weight, even if the mass or volume of the food is the same. I think the authors should revise these sentences to make it clear that it is the type of LULC change that is driving the change in ESV over time and space.

Minor grammatical/ typographical edits:

Line 20: "corrected the ESV had by the..." should probably be "corrected the ESV by the..."

Line 39: "Land use and land cover (LULC) change is an important forces..." should probably be "Land use and land cover (LULC) change are important forces"

Line 173: "has an impact service supply" should probably be "has an impact on service supply"

Line 404: "a large number of farmland" should probably be "a large amount of farmland" or "large areas of farmland"

Line 406: "the ESV decreased caused by" should probably be "the ESV decrease was caused by"

Line 436: "may led to" was probably intended as "may lead to" here

Author Response

Dear reviewer :

Thank you very much for your valuable feedback that we have used to improve the quality of our manuscript. We had carefully checked the manuscript and revised it based on your opinions. All changes in the article are highlighted in the marked manuscript. Point by point response to the reviewers’ comments are listed in response letter. The reviewer comments are listed in black text and our responses are given in red text.

Thanks again for the precious comments from you.

Best regards,
   Sincerely yours

 Yajing shao

Author Response File: Author Response.pdf

Round 3

Reviewer 2 Report

The authors have added some of the necessary details that were omitted in prior drafts. The Costanza et al. (2014) paper and the Bryan et al. (2018) seem to establish the practice of relative weights of ecosystem services in the BTM method; *however,* they do not alone justify the use of a sole valuation (crop production). Therefore, I'm glad to see that the authors have provided a theoretical foundation for this approach and that there is at least one prior study (Tianhong et al. 2010) that uses the 1/7 of crop production as the basis for the unit equivalent coefficient. This theoretical approach remains, to my mind, overly simplistic, but it is now described in adequate detail for readers to judge on their own.

I strongly encourage the authors to reconsider the stated aims and merit of the study. I believe that much of what they write about "spatial associations" is not actually relevant to the results of this paper. The authors, in their Response, write: "According to the calculation results of the GWR model, the impact of land use changes on the value of ecological services has positive and negative, which shows that the conversion speed and direction between land use types will have an impact on the value of ecological services." It seems we can agree that the different directions of the estimated effects on ESV (in the regression models) is due to the different modes of LULC change: conversion from a low-ESV mix of LULC types by area to a high-ESV mix of LULC types by area would present a positive coefficient, where the reverse would present a negative coefficient. The GWR model allows the authors to estimate different coefficients (positive or negative or not significant) across space. But does this "reveal the spatial association between LULC change and ESV"? No, because ESV change can be reduced almost completely (absent NDVI data) to LULC change. So, the results indicate that LULC change is heterogeneous across the study area.

Similarly, the authors write (lines 106-108): "LULC change has obvious spatial externalit[ies] resulting from place-based proximity, which has a significant impact on the ESV of adjacent land." Yet there is nothing in the methodological approach that accounts for these spatial externalities or spillovers. For example, a regression model with a spatially lagged term *would* formally model spatial spillover. A GWR by itself does *not* model spatial spillover; it models heterogeneity. As the results of this paper address neither spatial spillovers nor a *spatial* relationship between LULC change and ESV change, the authors should revise the stated aims and merit of the study.

Comments:

Lines 38-39: I think the phrase "Ecosystem services mainly include..." should just be "Ecosystem services include..."

Line 41: I don't think the phrase "inappropriate land use activities" is rigorous. The word used here was "improper" in a previous draft. Both "improper" and "inappropriate" are normative but the exact normative framing is unclear. A normative framing is unnecessary here. I think the authors need to either say something concrete, e.g., "Some land use activities have deleterious effects on ESV..." Or, better yet, just change "Inappropriate land use activities" to simply "Some land use activities..."

Lines 56-77: The new background material is helpful, but it may fit better in a subsection on "The Benefits Transfer Method," within the Methods section.

Lines 106-108: The authors write: "LULC change has obvious spatial externalit[ies] resulting from place-based proximity, which has a significant impact on the ESV of adjacent land." Yet there is nothing in the methodological approach that accounts for these spatial externalities or spillovers. For example, a regression model with a spatially lagged term *would* formally model spatial spillover. A GWR by itself does *not* model spatial spillover; it models heterogeneity. The authors should probably delete this sentence because they do not address spatial spillovers.

Line 211: "and used similar LULC types as agents to calculate the value of ecological services..." I don't think "agent" is the right word to use here--I'm not sure what it means in this context. Can you use a different word instead?

Line 225: Please put the year of the Xie et al. citation here so we know which paper in particular you are referring to, e.g., "In the study of Xie et al. (2008)..."

Line 260: The authors have since explained what "Evolutionary environment" refers to in the Response to Reviewers: "Environmental evolution refers to the treatment of human waste or industrial waste produced by the ecosystem. For example, there are many recycling production modes in the agricultural system, which leads to the farmland ecosystem playing an important role in environmental purification." However, they should make this clear to *readers* as well. I suggest changing "evolutionary environment" to "waste recycling" because the "treatment of human waste" doesn't really have anything to do with evolution.

Lines 561-566: These details on the sample size of the regression models should be moved to the Methods section.

Minor typographicaly/ grammatical corrections:

Lines 25-26: "was expanded eastward" should probably be just "expanded eastward"

Line 39: "and cultural service" should be "and cultural services"

Line 50: "can been used" should probably be "can be used"

Lines 88-89: "ecological environment problems" should probably be just "ecological problems" or "environmental problems;" based on the sentence following this one, it should probably be "environmental problems"

Line 106: "externality" should be "externalities"

Lines 203-204: "In 2008, Xie proposed that..." should be "Xie (2008) proposed that..."

Lines 216-218: The authors write: "Although the developed land is mostly impermeable surface, it does not belong to the natural ecosystems and can provide few natural ecosystem services." The word "Although" at the beginning of a sentence typically indicates that the first part and the second part are in conflict, or the second part is surprising given the first part. I think the fact that "developed land is mostly impermeably surface" is not in conflict with the idea that most people would not consider urban ecosystems to be "natural." I think the authors should change "Although" to "Because" (as one possibility). It may be best to remove the word "natural" as well because it is highly subjective.

Lines 226-227: "per hectare of farmland average annual natural food production..." This is a confusing phrase. Please change to: "per hectare of a farmland's average, annual, natural food production."

Line 227: "which can been easily traceable" should be "which is easily traceable" or "which can be easily traced"

Line 538: "consumer produce" was probably intended as "consumer products"(?)

Author Response

Dear reviewer :

Thank you very much for your valuable feedback. We had carefully checked the manuscript and revised it based on your opinions. All changes in the article are highlighted in the marked manuscript. Point by point response to the reviewers’ comments are listed in response letter. The reviewer comments are listed in black text and our responses are given in red text.

Thanks again for the precious comments from you.

Best regards,
   Sincerely yours

 Yajing shao

Author Response File: Author Response.pdf

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Overall, I find it very difficult to determine the purpose and importance of this study. There are a number of methodological details that need to be clarified. Some of them may be as simple as using a consistent and accurate terminology throughout the paper (e.g., clarifying "single land use dynamic degree"). But some of the methodological issues are more complex. I find that key results are not believable in light of the details that have been shared (e.g., Figures 3 and 4). There may be a simple mistake in data processing or perhaps crucial details in the methods have been omitted. Simply put, I can't follow how the authors have arrived at their conclusions based on the details provided.

Moreover, if we take the results at face value, it's not clear what their importance is. The study is extremely descriptive, with statistics on the exact amount of land area that is converted between years. While potentially interesting, its immediate relevance for land change studies, non-market valuation, or sustainability more generally is unclear. I encourage the authors to carefully think about how Xi'an and the changes they have observed are instructive to a wider audience, such as the audience that reads Sustainability.

Comments specific to lines in the paper are below:

L24-25: "Construction land" should perhaps be called "developed land" or "urbanized land" for clarity? Consequently, "through the occupation of" would be better put as "through the conversion of..." or "due to the conversion of..."

L25-26: "ESV decreased by 85.96×106 CNY..." I have no idea what this means. Acronym CNY is not defined in the abstract. I think the authors need to find a better way of stating this result... Is this some monetary value? It should probably be expressed as a percentage change over the study period, instead. Either way, it would appear to be a combined value of multiple, heterogeneous ecosystem services... At this point in reading only the abstract it's not clear what those ecosystem services are or why they are lumped together.

L30-33: I think the final statement in the Abstract should be more compelling. The authors write "The research results will help..." but as the research is about to be published I think it is preferable to be able to say that the research results "do" (present tense) something. Overall, it is a fairly generic statement to end the Abstract with.

L45-46: What is the ecological overload index? I think this statistic needs to be better contextualized.

L46: What makes land use change "improper?" I think this could be stated more plainly, e.g., "land use changes that reduce the ability of local ecosystems to provide food, fiber, clean air, or clean water..."

L55-57: Starting with "The research methods mainly include..." In the absence of a full background/ literature review section, it would be helpful if you could provide one or two references of papers that use each of these methods.

L66-67: "with a large research scope" What does this mean? Perhaps this phrase could be deleted.

L54-87: These three paragraphs seem like extraneous detail for the Introduction. Perhaps they should be moved to a new "Background" or "Literature Review" section instead? The Introduction needs to present a more concise and compelling case for what knowledge gaps this paper addresses.

L95-97: I do not understand what your first research question is. "How did land use structure change" is confusing because "structure" can be read as a verb or as a noun in this context What does "structure" mean here? Are you asking "how did land use change" or "how did [prior] land use affect urbanization" or something else altogether? Also, questions 2 and 3 do not seem to be distinct; aren't they really the same question--how has land-use/land-cover change affected ecosystem service value? If "dynamics" of LULC are truly of interest, I think you should state some dynamics of interest and briefly hypothesize how they will relate to changes in ESV. For example, is it "cyclic" LULC that is of interest? Or is it the rate (speed) of change that is of interest? These are dynamics that go beyond mere land-use/land-cover class conversions.

L105-115: Some of these details seem unnecessary. Instead of talking about soil colors or soil types, it would be more helpful to hear about how Xi'an is integrated into the wider national economy. This will help provide context for the "rapid" development and "population pressure[s]" described in the Introduction. Some things that I'd like to hear about in this section: Is there a lot of migration into Xi'an? Is that primarily for farm work or is it migration to major cities? How does Xi'an agriculture compare to the rest of Chinese agriculture, e.g., what fraction of production? These facts are more pertinent to land-use and land-cover changes than the soil color. You do have statistics on urbanization in the following paragraph, but the rural growth and rural production (agriculture) statistics would be a helpful addition.

Figure 1: The labels of provinces are quite small and hard to read as they overlap with one another and with dark border lines.

L134: "the interpretation accuracy of remote sensing images reached 90% in the five periods." Does this mean a minimum classification accuracy of 90% was achieved in each period? Perhaps it would be better to provide a range of classification accuracies, e.g., "classification accuracy in each of the 5 years ranged from a minimum of 85% (in year X) to a maximum of 95% (in year Y)."

L141-142: "By comparing the actual differences between grid areas of different grid sizes, the optimal grid size of this study was determined to be 1x1 km..." How exactly was this determination made? Using a variogram?

L155: There should be ellipses (...) on the fourth row of this matrix as well, to implicitly represent the land-use types 3 through N on the vertical axis. Alternatively, a more succinct representation of this matrix (in TeX format) is: $[S_{ij}]$. The description should also be re-written for clarity, e.g., "where the i,j element represents the number of cells with land use i in Year 1 and land use j in Year 2." The phrase "observation period i to j" is misleading because i and j index land-use types, not periods. It is the axes which represent two periods.

L158-169: What is the difference between "single land use dynamic degree" and "comprehensive dynamic degree"? Is this a typographical error? On Line 165 you write: "where K_i is the dynamic degree of land use type i;" however, K_i is, in fact, simply the rate of change or "change rate." Is this also a typographical error? Or are you saying that "single land use dynamic degree" and "change rate" are equivalent? If so, I would strongly recommend just calling it "rate of change" (or "change rate") as I have never before heard the phrase "single land use dynamic degree" and it is easily confused with whatever the "comprehensive dynamic degree" is.

Line 169: Why is the sum in the denominator multiplied by 2?

L174-186: Much important detail seems to be coming from the references numbered 33 and 34. Instead of forcing the reader to look up these papers, some key details should be provided in this paper, particularly the " ecosystem service equivalent scale" (as a table?) Is this what Table 1 contains? You need to explicitly reference Table 1 here and provide some more background information. Where, in general, do these valuations come from (beyond the work of Xie et al.); from a review of the literature? Or from an expert survey? Similarly, what makes the agricultural production value the "equivalent" of other values the land could provide? This is unclear and on first read, without reference to Table 1, it almost seems like you are saying that the only ecosystem value that exists is agricultural production.

Line 181: The term "M" in Equation 4 is undefined.

Line 185-186: You write: "The economic value of unit ecological services in Xi'an is calculated to be..." What are "unit ecological services" and why is a fixed value required? Or is this a statistic, e.g., the mean value for Xi'an???

L192-194: We're just now getting to how land uses are defined. These details--on how "construction land" is defined, in particular, should be moved up earlier in the Methods section.

L194-195: You write: "According to existing research results, the ESV value of construction land is 0." Why should it be zero? "Construction land" (again, a better description might be "urbanized" or "developed" land) certainly provides value to humans. Is it zero because we assume that developed land is not part of an ecosystem? There are clearly many assumptions to this ESV framework. The authors are relying on the Xie et al. reference(s) for much of the detail about this framework but Sustainability is a journal for a broad audience, so much more detail and motivation about ESV needs to be included in this paper.

Line 211: The way this equation (Equation 7) is formulated, the "change rate" will always be positive. While it may be desirable to have an absolute rate of change available, the direction of change is also important; e.g., has the ESV provided by forested lands increased or decreased? Later, you show a figure (Figure 4) that implies this absolute rate of change can be calculated in each year... Which begs the question, what are you calculating the maximum and minimum over? If every parcel with the same land use and the same service type has the same value every year, the only way that total ESV by land use can change is if the area of that land use changes. However, Equation 7 suggests that the maximum/ minimum ESV is calculated within a land-use class (i.e., across time) while Figure 4 implies that it is calculated within land-use class and year (across space *and* time). Since a spatial maximum/ minumum would be the same in every year, Figure 4 just doesn't make sense.

L214-226: What are the independent variables in this regression? What is the dependent variable, for that matter??? Is Y_i the observed land-cover/ land-use class (in a given year)? There are not time subscripts, either... This is totally unclear, based on any stated research objective or research question from earlier in the paper. The independent variables are not described here, either. What is the point of doing GWR? This paragraph serves to describe GWR as an abstract method, but what are we trying to learn???

Line 220: "the results are more in line with the objective reality..." The GWR's assumption that the regression coefficient is not constant across space may be important for studies of ecological service value, but in other studies a global regression coefficient might be more in line with "objective reality." I would suggest restating this sentence to narrow this claim to studies of ecological service value, such as the paper cited as reference #36.

L237-247: And Table 3... You write earlier "The comprehensive dynamic degree reflects the comprehensive change rate of all land use types..." And Equation 3 returns a single value across all land-use types (i.e., the quantity "LC" is not indexed by i) Yet Table 3 has a row for *each* land-use type. Is Table 3 really a listing of the "single land use dynamic degree" (L159)? Again, I think the phrase "single land use dynamic degree" should be replaced with either "[land-use] change rate" or "rate of [land-use] change."

L274-275: Figure 2 needs to be made much larger in the final version. In particular, the legend and scale are too small to read.

L285: I don't understand the notation "333.08×106 CNY". I see that "CNY" is a financial label for money in renminbi, but I don't understand the "x" or multiplier symbol. Is this some kind of explicit reminder of inflation? I think it should be stated that the values are, e.g., "2018 CNY," that is, values as of the year 2018 or adjusted to match 2018 prices.

L295-296: "Because of the implementation of divination balance policy..." What is "divination balance policy"? This is the only use of this phrase throughout the paper It strikes me that there is not enough background information on what makes Xi'an an interesting place for what seems to be a very descriptive study. Is "divination balance policy" one of the more interesting socio-political aspects of this study area? I strongly recommend the authors put together a more thorough Background section that links the descriptive results of this study (e.g., amount of area converted between land uses), which are not very meaningful on their own to wider audience, and the social, economic, and political drivers of land-use change in Xi'an, China, and the world.

L298: Table 5... The ESV by land use does not change much at all throughout the study period. Perhaps this table is unnecessary? Furthermore, if the value of a given ecosystem service in a given land use is fixed throughout the study period, this table provides no new information--the total ESV by land use in a given year is just a product of the change in land area and the fixed value the land provides.

L302-323: And Figure 3... How is it that lands in three counties in Northeast Xi'an can go from >194M CNY to <10.39M CNY in five years? And in subsequent years their lands never provide more than 22M CNY? Since the ecological service values of each land-use type are fixed, the only way that a given parcel's ESV can change is if its land use changes. Yet the top of Figure 3 would seem to imply every cell in these three counties has changed land use between study years. It is totally unclear how ESV could change so dramatically in this way without dramatic land-use change, which the authors' analysis has revealed to be largely characterized by outgrowth of urbanized areas. In Lines 315-319, the authors note that ESV increases in the "southern areas," but it's not at all clear why this would be so. The analysis of Figure 3 is made all the more difficult in that it uses a diverging color scale (blue to red) for a monotonic quantity (money from zero to infinity) and because the range of the color bins changes between each subplot. I'm guessing these are quintiles (5-quantiles), which is reasonable for exploratory analysis of a figure, but not suitable for comparison across different data subsets.

L336-337: As I noted earlier, Re: Figure 4... Equation 7 suggests that the maximum/ minimum ESV is calculated within a land-use class (i.e., across time) while Figure 4 implies that it is calculated within land-use class and year (across space *and* time). Since a spatial maximum/ minumum would be the same in every year, Figure 4 just doesn't make sense in light of Equation 7.

----------------------------------------------------------

Minor typographical/ grammatical edits are below. These do not require any response in a "Response to the Reviewers" document; they are provided purely for the benefit of the authors.

Throughout the paper, I think that in-line citations should have a space between the word and the brackets, i.e., on line 68, "national[15,16]" should instead be "national [15, 16]."

L64-65: "the contradiction between ecological environment and land use rapidly increased." I think this would be more clear as "the conflicts between ecosystem service provision and land use are increasing."

L69: "as evaluation unit" should be "as the evaluation unit."

L71: "ecological services and its" should be "ecological services and their"

L86-87: This sentence is confusing as written... I think "Therefore, research on...is slightly insufficient" was perhaps meant to be something like, "Therefore, more research on [these topics] is needed."

L188: "equivalents" should be "equivalent"

L215-217: Two sentences starting "GWR model..." should probably be either "The GWR model..." or "A GWR model..." It sounds awkard without an article like "the" or "a".

L247: In Table 3, it would be helpful to visually distinguish, if possible, the final column labeled "2000-2018" as this column breaks the temporal ordering of the others to its left. Perhaps a different shade/ background color?

L264-265: "urbanization development" should probably just be either "urbanization" or "development". Or perhaps you meant "urban development"?

L267: "a large number of cultivated land and woodland are..." should be "large amounts of cultivated land and woodland are..."

L270-271: "The economic development of southern area" should be "The economic development of the southern area..."

L270-271: "The economic development of southern area was slower than that of other regions, but the ecological environment was better." What does it mean that "the ecological environment was better"? Perhaps it is sufficient to say that the lands are less developed.

L273: In Table 4, it seems that the forward slash character "/" is used to denote "not available" or "no data" but I would strongly recommend a different character be used, such as an em-dash or long dash.

L303: There are two periods ("..") after "(Figure 3)"

L312-313: "a large number of" should probably be, e.g., "a large amount of..."

Reviewer 2 Report

The study uses satellite images to monitor land use and land cover change within an 18-year period in Xi’an (China) and develops a GWR model to link such change with ecosystem service values in a spatially explicit manner. An interesting finding is the nonlinear association between LULC change and ESV, which may have profound implications for land planning and ecosystem conservation. I believe the findings would contribute to the existing literature with theoretical understanding while adding to policy practices with empirical knowledge. Given its merit, the paper with a revision would be more rigorously effective in presenting the research ideas and methods. Please see below my detailed comments and suggestions, most of which aim to encourage some clarifications with an emphasis on methods.

Line 19, dynamics include the meaning of change, so use the dynamics of LULC or LULC change.

Line 25, superscript the power of 6.

Line 91, by a GWR model.

Line 134, how was the Kappa coefficient calculated? If it is from the original data source, a citation is needed; if done by the authors, details on reaching the accuracy need to be provided (probably in Appendix), such as sampling for validation, accuracy for each year, confusion matrix, etc.

Line 141, the process of “comparing the actual difference” is rather vague for determining the so-called “optimal” grid scale. Please provide more information on the criteria of choosing the proper scale.

Line 156, may add “in this study, n = 6.”

Line 170, it needs to add a notation for LC. Is it Land Comprehensivity, Comprehensiveness, or else?  

Line 181, the symbol of m is double used with one sown area and the other total number of crops, raising confusion; I think the m above the sigma symbol should be n. M is not denoted, being likely the total sown area or cultivated area. It also needs justification for the 1/7 adjustments, as Xie et al (2015) argued the roughly estimated ratio of ES equivalent coefficient to per-unit grain yield is 1/7.

Line 184, are these temporally averaged values, i.e., the sum of the values of the five years divided by five? What is the point of calculating this aggregated value, given the time-series data?

Line 185, what is the CPI value? Is it assumed fixed each year?

Line 195, this part of construction land (including urban public mining, residential, unused, bare lands) is a bit confusing. Does the cited paper explicitly set the entire construction land category’s value to zero in methodology? It appears some unutilized land still grabs a bit (although trivial) values, which, also shown in the table, is somehow inconsistent with the text. Additionally, it would be more convincing if citing 1-2 more papers of application with clear statements such as Wang et al (2018).

Line 222, key information is missing for the GWR. For instance, what are the independent variables? How were sample selected? How was the GWR calibrated with optimal weights for samples? Moreover, it would be more rigorous if citing papers such as Fotheringham et al. (2001 & 2003). 

Line 486, some of the text are not limitations, such as ESV scale, CPI, grid, GWR. These parts should be removed or integrated in other sections.

Line 619, reference are incorrectly aligned, making it difficult to referring them to citations in the main text.

References

Fotheringham, A. S., Charlton, M. E., & Brunsdon, C. (2001). Spatial variations in school performance: a local analysis using geographically weighted regression. Geographical and Environmental Modelling, 5(1), 43-66.

Fotheringham, A. S., Brunsdon, C., & Charlton, M. (2003). Geographically weighted regression: the analysis of spatially varying relationships. John Wiley & Sons.

Xie, G. D., Zhang, C. X., Zhang, L. M., Chen, W. H., & Li, S. M. (2015). Improvement of the evaluation method for ecosystem service value based on per unit area. J. Nat. Resour, 30(8), 1243-1254.

Wang, Y., Li, X., Zhang, Q., Li, J., & Zhou, X. (2018). Projections of future land use changes: Multiple scenarios-based impacts analysis on ecosystem services for Wuhan city, China. Ecological indicators, 94, 430-445.