Ecosystem Service Responses to Land Use Change in Southern Guangzhou—The Practice of Applying Natural Resources Integrated Database for Research

Round 1

Reviewer 1 Report

It looks better now. However, there are some minor suggestions that the abstract should be simplified. The current version is still too long. It is only necessary to explain the core contributions of this paper. I recommend limiting it to 250 words.

Author Response

Thanks for your opinion. We have simplified the abstract, and the body content of the current version is about 250 words.

Author Response File: Author Response.docx

Reviewer 2 Report

Dear authors

Thank you for giving me a second opportunity to read your work. The manuscript “Ecosystem service responses to land use change in southern Guangzhou: The practice of applying Natural Resources Integrated Database for research.” is very interesting as I stated before. This version has much improvements and denotes hard work from the authors, Nevertheless I still have some questions/doubts and suggestions.

What is the Land use classification error? Without knowing the accuracy of LULC maps it is not possible to say the importance of the changes since the error is cumulative. If we have an accuracy of 92% in two years classifications, the time accuracy is 0.92 x 0.92 = 0.85 (85%). Thus, changes below 15% can be completely justified by classification errors.

Moran′s indexes are well known and do not need to be explained in detailed. A Reference is enough (Anselin).

Line 348 – Figure 2 – Please do not name as legend as legend (use LULC for example). We cannot read the legend. One cannot start a section with a figure. Figure and tables must be referred in the text before their appearance.

Natural Breaks (Jenks) is a bad method to compare different maps

Figure 6 shows that the ESD scatter points in 2004, 2010 and 2016 were mainly distributed in the upper-right and lower-left quadrants, while the scatter points in the lower-right and upper-left quadrants were relatively small. Where are the dates and the quadrants in the figure?

Figure 9 a R2 of 0.6 Is good?

There is no discussion section

As minor comments, I will stress:

I think this in the title “u——T” is a mistake

hm are not in italic and there must be a space between the number and the units

Some figures are not referred in the text. They have to be mentioned before their appearance.

please replace “formulas” by equations or expressions. Formulas are usually reserved to chemistry

Figure 8 is unreadable

The X axis indicates the east direction, the Y axis  indicates the north direction, and the Z value represents the ESD”. This goes to the title of the figure not within the text

References are not according to the journal requirements (numbered brackets)

Best regards

Author Response

Thanks for your opinion. We have adjusted the paper in following:

The land use classification error means that the land use classification results do not match the definition in Table 1. We carried out random sampling through human-computer interaction, and the sampling rate is about 8% of the total number of land use patches (the land use patches are both closely to 60,000, and the workload of comparing and checking for each patch one by one is too large). The results of land use classification in 2010 and 2016 was superimposed and analyzed with the previous year, and the patches with changes in land use types will be focused on spot checks. Patches are corrected locally if land-use classification errors are found, which make the total accuracy of land use classification better. Although classification errors are unavoidable, after the above process, it was controlled.

Although Moran's is well known in geography field, many literatures still retain explanations. I think it is necessary to reserve the explanation, and I hope to keep it in the paper.

"Legend" has been removed. But in fact, I can't understand what's the meaning by "do not name as legend as legend". The figures and tables in the paper have been adjusted to below the text.

We changed the classification method in figure 3 to the equal-spacing classification method, with 10 as the divided interval. Although the classification of some local grid has changed, the overall trend of the spatial distribution of each classification has hardly changed.

In the figure 3, we supplemented the annotation of quadrants (on the left figure) and year. Regarding the quadrants in the figure, it is the 4 regions divided by the horizontal axis and the vertical axis in the plane rectangular coordinate system. The horizontal axis distinguishes left and right, and the vertical axis distinguishes lower and upper.

For a linear trend line in the scatter plot, the R2 value of 0.62 is not low (the absolute value of R is close to 0.79). The regression equation of this trend line can explain 62% of 1Km × 1Km grids that built-up increase cause to ecosystem services reduced. This is a good fitting degree, which can reflect the negative trend of urbanization on ecosystem services.

Regarding to the title, could we modify the title as “Ecosystem service responses to land use change in southern Guangzhou: the practice of applying Natural Resources Integrated Database for research”? After our discussion, this title is a compromise one.

We adjusted other expressions: the italic in units were canceled, the spaces between the number and the units were added, the “formulas” were replaced by equations, the X, Y axis were added on the title of figure 4, and the references were modified as numbered brackets.

Author Response File: Author Response.docx

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

This paper analysed the LULC and ESV changes in southern Guangzhou of China and then proposed a regionalized and interannual ESV correction method. This study used several data sources to improve the ESV and the result of the computation were published as Web services and displayed on the Visual Web Front-End. The paper is well written and structured. Following minor corrections may be addressed before publishing the paper.
• Figure: A well-prepared map is necessary showing the boundaries, legend, north arrow, and scale
• Section 2.2: the authors mentioned that they reclassified the LULC. I am not clear whether they collected the data from the secondary source or they created/classified LULC themselves. If the authors created the LULC, then please describe the methodology
• Figure 2 needs to be improved following the comments for Figure 1.
• Table 3: If the authors want to present the change of LULC, they need to present the time of year as duration like 2004-2010 in the heading. A single point of time does not reflect the change
• Figure 3: the legends should be improved to make them clearly readable
• Figure 7, 8: Please improve the legends

Author Response

Dear reviewer: Thank you for your revision. We have remade the figures in this paper and supplemented map factors such as legend, north arrow, and scale.

Regarding to LULC, we have revised the expression of the full text to land use, and explained the source and classification method of the data in the 2.2 Land Use Category:

The land use data in this paper is based on the scientific research data from Guangzhou University. These land use data are supervised by Support Vector Machine(SVM) from Landsat 5 Thematic Mapper(TM) images in 2004 and 2010 and Landsat 8 Operational Land Imager(OLI) images in 2016. The land use types are reclassed into 8 types, and the classification standard according to “the Contents and Indicators of Basic Geographic Monitoring (GQJC03-2017)” issued by the State Bureau of Surveying and Mapping Geographic Information of China. After partial correction of SVM supervised classification results, the accuracy of random sampling of land use patches are better than 90%.

Reviewer 2 Report

In this paper, the author takes the response between ecosystem services and land use change as the research topic, which is a very interesting topic, especially for the current rapid urbanization in China, has very important significance. However, as far as this paper is concerned, there is still a long way to go. The whole paper is more of a method-oriented scientific report than a journal paper. Specifically:
1. The current content of the abstract is too long and lacks corresponding refinement and summary. It is suggested to list only some key research logics and findings without showing too many details.
2. At the beginning of Introduction, the realistic background and theoretical research status of ecosystem services and land use change should be explained. However, at present, the structure is too loose to allow readers to see the introduction of your research gap. Secondly, the third paragraph explains the evaluation method for studying the value of ecosystem services, which is not suitable for this part, but should be included in the literature review or methods section. Next, Why should we study the spatial correlation and heterogeneity of geographical objects? The reader is not convinced. The fifth and sixth paragraphs introduce the NRIDB database, but why this? Is this the source of the data? If it is, it would be better to put it in the Data section, not here. The seventh paragraph introduces the study area. Would it be more appropriate to move directly to the overview of the study area in the next section? On the whole, the Introduction needs a lot of modification at present to conform to the elements of a scientific paper, which is the starting point of the whole research and must be improved.
3. The title of the second section, "Method", seems to be inaccurate. According to its content, not only method, but also study area and data, the current title is not accurate. The current subtitle directory is also not appropriate, which is generally juxtaposed with study area, Method and data. The current organizational structure is a bit chaotic. However, in the overview of the research area, the drawing in Figure 1 is not standardized. On the one hand, it is suggested to draw the map of the research area in a normative and standard way. On the other hand, the scope of the research area in the current map is not clear enough for readers to see clearly. Section 2.3 is about the introduction of ESV estimation methods. The current content are too long. It is suggested that we only introduce the principle and usage of the method instead of such a long introduction. In section 2.4 grid Analysis, this is not so much a method as a research object or data processing mode, according to the introduction. Sections 2.5 and 2.6 partially introduce some methods of spatial autocorrelation analysis, but these are not the author's original work, so it is suggested to supplement the corresponding citation sources. In Section 2.7, the word "data consolidation" is not accurate. At present, it is only the data source, and the existing data are not cleaned and sorted.
4. In Figure 2, it is suggested to adopt a unified style for the three pictures, but the current style cannot be well understood by readers.
5. Section 3.4 analyzes the spatial distribution of ESD, but why different grid resolutions are used for this analysis? It doesn't seem clear. And the reader doesn't see the value in this approach.
6. In Section 3.8, simple linear regression analysis is used to analyze the impact of urbanization on ES, but the data scatter diagram shows that the relationship is not so simple.
7. Section 3.9 is a bit obtrusive in the text, why is it added? Does the research in this paper seem irrelevant?
8. The discussion section of Section 4 is too long, so it should be refined and summarized instead of simply repeating the previous content.
9. Similar to the previous point, the difference between Conclusion and Discussion is not clear, so much content should be included in discussion. Conclusion should be a summary and outlook of the article.
10. The style of the reference is not standard, so it is recommended to modify it properly.

Author Response

Dear reviewer: Thank you for your revision, we have made a series of modifications to manuscript. The specific explanations are as follows:

1. We have rewritten the abstract
2. Paragraph 2 of the original manuscript is a further explanation of the research background of this paper. I have rewritten and integrated it to the new 1st paragraph to describe the research background and necessity of this paper.

The 3rd paragraph of the manuscript is an overview of the research status of the equivalent factor method, and the necessity of ESV correction for regionalization and inter-annualization in this paper. We supplement the literature review in the new 2nd paragraph.

After evaluating the spatial distribution of ESVs in the study area, it is necessary to further explore the clustering patterns and association characteristics of the spatial distribution of ESVs through spatial correlation and heterogeneity studies. This is helpful for us to understand and further study the spatial pattern characteristics and driving mechanism of ESV in the process of urbanization in this region.

The new paragraph 4 is a general introduction to the research and process carried out in this paper.

NRIDB is the GIS applied in this paper for data collection, spatial analysis computing, and map visualization. The use of governmental GIS for spatial analysis computing of ecosystem services and the display of the results on the Web is one of the innovations of this paper, but also a response to the theme of "Applications of Remote Sensing and GIS Integration in Natural Resources and Environmental Science" Special Issue. A new paragraph 5 therefore introduces the NRIDB and how the NRIDB can be used to carry out the research in this paper. Regarding the source and classification method of the land use data in this article, we have revised 2.2 Land Use Category and explained it.

3. We modified 2nd section to "Study Area, Data, and Methodology", and adjusted the order of each part.

The correction method of ESV coefficient is one of the key points of this paper, and it is necessary to explain it clearly.

Figure 1 has been remapped, and we supplemented map factors such as legend, north arrow, and scale.

The original 2.4 grid analysis, has been revised to 2.5 grid processing.

After reviewing the relevant references, we supplemented the annotated citations in 2.5 and 2.6.

The original 2.7 was modified as a data source and adjusted to 2.5

4. The maps in Figure 2 has been modified into a unified style.
5. Global spatial autocorrelation for ESD distribution in 3.6 is based on ESD of various scale grids in 3.4. Considering that the 5 groups of maps occupy a large space and the showing value is not prominent, we only retain the 200m-scale grid to display the spatial distribution characteristics of ESD.
6. At present, only the linear relationship that the scatter plot expresses the negative trend is directly displayed. The actual situation will be more complicated. However, the impact mechanism of urbanization on ecosystem services is the direction of our research in the next stage.
7. There is no 3.9 in the manuscript

8.-10. The conclusion and discussion have been revised and some contents have been deleted. The style of references has also been modified accordingly.

Reviewer 3 Report

Dear authors

Thank you for giving me the opportunity to read your work. First of all I am sorry for taking so long to review it. The manuscript “Ecosystem service responses to land use change in southern Guangzhou: The practice of applying Natural Resources Integrated Database for research.” is very interesting. Nevertheless I have some questions/doubts and suggestions. Mainly, I just want to start some discussion on specific points and clear some ideas.

Line 57 – Caution - Changes in land use patterns do not necessarily correspond to changes in land cover patterns

There is a huge gap between lines 86 and 87. One need a link between the necessity to modify Xie’s equivalent factor table and spatial correlation and spatial heterogeneity.

Line 102 – “Few studies have”, references needed

Line 311 – “The local Moran’s I index (LISA)”. This is a common misunderstood. LISA stands for Local Spatial Autocorrelation Analysis. Local Moran’s I index is a LISA method but LISA is much more than local Moran’s I index. Therefore, one cannot use LISA as acronyms for local Moran’s I index.

Lines 110 to 134 have no relevance to the paper.

Line 192 – “In order to comply with the management regulations and relevant laws of government data, the LULC data in this paper adopted the scientific research data compiled and  provided by Guangzhou University. We need references, how the data is acquired? What methods are uses? What is the classification error? And so on.

Line 281 – grid matrixes are by definition of the same size and non-overlapping

Line 295 - Moran′s indexes are well known and do not need to be explained in detailed. A Reference is enough (Anselin).

Line 348 – Figure 2 – Please do not name as legend as legend (use LULC for example). We cannot read the legend. One cannot start a section with a figure. Figure and tables must be referred in the text before their appearance.

Line 351 table 3 – 1.28 rounds to 1.3 and not 1.2. Also 1.28 is the 12 years change divided by 12 and not the yearly change rate. The yearly change rate is progressive since the % of the second year is calculated upon the initial area plus the change area of the first year. Without knowing the accuracy of LULC maps it is not possible to say the importance of these values.

Line 471 - Natural Breaks (Jenks) is a bad method to compare different maps

Line 474 to 480 – the text is analyzing figure 4 without referring to it and this figure is in the next section

Line 481 to 487 has no relevance to the work.

Line 543 – “first and third quadrants, while the scatter points in the second 543 and fourth quadrants are relatively small” first, second, third and fourth quadrants?? Where is this in the figure?

Line 584 – with a R2 of 0.6. Is this good?

Line 591 – section 3.9 has no scientific relevance is just technology description. Moreover it as no interference in the paper objectives and results.

There is no discussion section

As minor comments I will stress:

I think this in the title “u——T” is a mistake

Line 7 - 510006), the ) is out of space

Line 10 – LULC was not previously defined

Line 136 - LULC was not previously defined. It has to be both in abstract and in the main text

Line 170 – km are not in italic and there must be a space between the number and the units

Figure 1 is not referred in the text. It has to be mentioned before its appearance. Also this is not a proper image, it needs scale and a world context.

Line 187 – please replace “reclassed” by reclassified

Line 200 – Table 2 reference is  Xie et al. (2015)? If yes please add it.

Line 287 – please remove “number of non-overlapping”

Line 289 – plese remove “and a series of square grid matrixes are constructed from these square grids as geospatial statistical unit.”

Line 407 – please replace “formulas” by equations or expressions. Formulas are usually reserved to chemistry

Figure 3 legends are unreadable. There is no A, B, C, ……

Line 494 – “The X axis indicates the east direction, the Y axis 494 indicates the north direction, and the Z value represents the ESD”. This goes to the title of the figure

Line 531 - Therefore, this paper (uses??)  Moran scatter plots and LISA

References are not according to the journal requirements (numbered brackets)

Best regards

Author Response

Dear reviewer: Thank you for your revision, we have made a series of modifications to manuscript. The specific explanations are as follows:

Line 57, 192, 348, 10 and 136：

Regarding to LULC, we have revised the expression of the full text to land use, and explained the source and classification method of the data in the 2.2 Land Use Category.

The land use data in this paper is based on the scientific research data from Guangzhou University. These land use data are supervised by Support Vector Machine(SVM) from Landsat 5 Thematic Mapper(TM) images in 2004 and 2010 and Landsat 8 Operational Land Imager(OLI) images in 2016. The land use types are reclassed into 8 types, and the classification standard according to “the Contents and Indicators of Basic Geographic Monitoring (GQJC03-2017)” issued by the State Bureau of Surveying and Mapping Geographic Information of China. After partial correction of SVM supervised classification results, the accuracy of random sampling of land use patches are better than 90%.

LRTCG is a government system. It is inappropriate for the official version of governmental data to be made into a map and published publicly. Therefore, we decided to use the scientific research data.

Line 102：

We found relevant references, and supplementary descriptions and annotated citation references in the text.

Line311：

The expression in the full text has been modified to "local spatial autocorrelation analysis".

Line 110-134：

This part introduces the background of NRIDB application in ecosystem services research in this paper.

NRIDB is the GIS applied in this paper for data collection, spatial analysis computing, and map visualization. The use of governmental GIS for spatial analysis computing of ecosystem services and the display of the results on the Web is one of the innovations of this paper, but also a response to the theme of "Applications of Remote Sensing and GIS Integration in Natural Resources and Environmental Science" Special Issue.

Line 295：

We supplemented the annotated citations of Anselin in manuscript.

Line 351：

The analysis in manuscript is based on the trust that the "data is accurate". 2.2 in the manuscript has supplemented the description of the accuracy of the data.

The calculation method of Table 3. is correct. Taking Built-up as an example, the calculation logic is as follows

40518.61/35108≈1.1541: The built-up area in 2016 is 1.1541 times that in 2004.

(1.1541)^(1/12)≈1.012: During 2004-2016, the annual built-up area was 1.012 times that of the previous year, which was converted into an average annual increase of 1.2%. That is, the built-up area in the new year increased by 1.2% over the previous year.

Checking calculation with 4 decimal places: 1.012^12≈1.1541.

Line 471:

Jenks classifies the research objects into groups with similar properties, which is statistically significant.

Jenks determines the best arrangement of values in a group by iteratively comparing the mean value of each group and the elements in the group, and the sum of the squared differences between them and the observed value. Then breakpoints in the ordered distribution that minimize the sum of square differences within the group. Compared with the method of evenly dividing numerical intervals, Jenks' breakpoint itself is a good boundary of classification, and the numerical agglomeration characteristics of each classification interval are more obvious.

In future research, we will consider other breakpoint methods.

Line 474-480:

This part analyzes the change characteristics of specify areas in Figure 3. The last few sentences have been deleted.

Line 481-487

This part is to explain the computing process of data by using NRIDB. Reflect the role of NRIDB in manuscript. I would hope to retain this part of the description in manuscript.

Line 543

This refers to each partition in the scatterplot of Figure 6. It is inappropriate to refer to "quadrants" in the manuscript. Quadrants 1-4 have been modified to upper-right, upper-left, lower-right and lower-left partitions.

Line 584

An R2 of 0.6 means that the absolute value of R is close to 0.775. The linear correlation between the independent variable and the dependent variable is very good.

Line 591

There is no 3.9 in the manuscript

Line 7

Brackets have been removed.

Line 170

We supplemented a space between the number and the units. Figure 1 has been remapped, and we supplemented map factors such as legend, north arrow, and scale.

Line 187, 200, 281, 287, 289, 494, 531

It has been modified according to the comments.

Line 407

It has been modified according to the comments. Figure 3 only retains grids with a scale of 200m.

Round 2

Reviewer 2 Report

Thanks for the detailed replies from authors. In general, authors have made great modifications and improvements in this version. Except for some small issues, it is basically acceptable:  
1 The abstract is still a bit too long, so it is recommended to simplify and highlight only the core contributions of this manuscript.  
2. In Figure 1, the authors have indeed made some improvements, but the text and symbols in the right figure are not clear at all. It is suggested to make the satellite base transparent, so as to highlight the labeling information.  
3 In Figure 5, the Moran's I values of different years and scales are put together, which does not seem to be very comparable, It is suggested to compare either in the year dimension alone or in the perspective of different scales.  

Author Response

Dear reviewer

Thank you for your revision. I made some improvements to the paper as follows:

1. Abstract has been simplified from 438 words to 337 words.
2. Figure 1 has been remapped. We changed the style and color of lines, adjusted the satellite image base map of the background, and the size and color of text labels on the right side of Figure 1. The scope of Guangdong Province and Guangdong-Hong Kong-Macao Greater Bay Area on the left side of Figure 1 has also been adjusted. The labels and colors of the remapped Figure 1 are highlighted. I hope that the size of Figure 1 can be reserved as width: 18.46cm & height: 8.11cm, so as to make full use of the space on the page and help readers to read the content of the figure more clearly.
3. Regarding Figure 5, the reviewer's revision is very reasonable. It turns out that the content in Figure 5 is indeed not easy to distinguish, and the comparability is not obvious. However, I still hope to express the spatial autocorrelation trend of ESD grids for various sizes in both spatial and temporal dimensions in one graph. Therefore, the Figure 5 design is optimized. ESD grid of the same size is displayed in the same color, in 2004 as a straight line, in 2010 as a dot, and in 2016 as a dash. In the adjusted Figure 5, we can know the change of spatial autocorrelation of ESD grid of each size and year with spatial scale horizontally. Vertically, we can compare the spatial autocorrelation differences of ESD grids of different sizes in the same year, as well as the same size over time.