Research on the Spatiotemporal Evolution of Mangrove Forests in the Hainan Island from 1991 to 2021 Based on SVM and Res-UNet Algorithms

Round 1

Reviewer 1 Report

The manuscript is very well written and falls within the scope of this Journal. Please address the following comments for improvements.

L2: Consider revising the title, which does not accurately represent the content of this manuscript. According to the purpose statement (L125-134), this study aims to improve the accuracy of remote sensing estimates of tropical mangrove forest spatial information using SVM machine learning and Res-UNet deep learning algorithms.

L145-146: Provide more detailed information on the field surveys. Did the author (or other investigators) visit all of the 660 sites within one month (i.e., Oct 2021)? What about the area size of each site? What and how did the author record during the field surveys?

L150: “A1” > “Table A1”

L151 (Also L191-192, L236, L484-485): “the visual interpretation of Google Earth imagery” > Since the ground truth used for the accuracy assessment is fully based on this visual interpretation, the author should provide more detailed information in the manuscript. What exactly is the “Google Earth imagery”? Which satellite imagery is it, what is the resolution, and when was it taken? How did the author conduct the visual interpretation? There might be a criticism that the interpretation can be subjective. How did the author ensure its objectivity?

L153 (Figure 1), L621 (Table A1): Explain the abbreviations of city/county names (e.g., “DZ”). “Aegiceras Corniculatum” > “Aegiceras corniculatum”

L153 (Figure 1), L427 (Figure 10): Use the colors that are distinguishable with each other.

L193: Clarify how the author divided samples into training and validation ones.

L216 (Figure 3): Explain the abbreviations of “conv” and “CAT” used in the figure.

L218 (Figure 4): Define “x” and “F(x)” used in the figure.

L278: “P_ij” > “p_ij”. Also define “i” and “j”.

L227: Also define “ɛ” in the text.

L294-297: “Eight indicators” > “Eight indicators of each city/county”? What about the information sources?

L300: “Machine Learning” > “SVM Machine Learning”? //Consider specifying the machine leaning method. In my understanding, deep learning is a type of machine learning methods.

L313 (Figure 5), L330 (Figure 6): The font in each map is too small to read.

L316 (Table 1), L333 (Table 2): How did the author get the validation samples between 1991-2015? According to L145-146, land cover type in each validation site was confirmed through the field survey in October 2021. Did the author use the same sites for the validation between 1991–2015? Then, how did the author confirm the land cover types? Why is NOT the total number of validation sites identical to “660” between 1991–2015?

L317: “Deep Learning” > “Res-UNet Deep Learning”?

L339: “Machine Learning and Deep Learning” > “SMV Machine Learning and Res-UNet Deep Learning”?

L349: “Figure 8b,c” > “Figure 8c,d”?

L353: Clarify the spatial scale. How long is the one side of a square? Was the validation site a square plot? Is the satellite images from Landsat (30 m resolution)? Which year was it taken? How did the author generate the mangrove forest area shown in “Ground-Truth”?

L364: “2233.80, and 3438.63 ha” > Explain in the Discussion chapter why there was such a large increase in mangrove forest crown cover within 6 years between 2015–2021. One reason might be “the increase in planted mangrove forests” (L518), while it might not affect significantly because “the survival rate of artificially planted mangrove forests tends to be very low” (L520-521). Since the mangrove forests area in 2015 estimated by Res-UNet was rather small compared to SVM and previous studies (Table 8), readers might suspect if it is fairly accurate.

L381 (Table 3), L396 (Table 4): I found no values in these tables. For Table 3. it should be “0” if no area was classified as mangrove forest crown cover. Then, all values can be calculated in Table 4.

L394: “Lingshui Li Autonomous County increased between 1996 and 2010” > In Table 4, Lingshui Li Autonomous County has no value in 1996–2000.

L400: “a general continuous increase during 1991–2021” > This is not correct, since NP is decreased in 1996 and 2010.

L411-414: “In summary, …” > This sentence should move to the Discussion chapter, since it contains the author’s interpretation of derived results.

L439 (Figure 11), L441 (Table 6): > Show the sources of the statistical information.

L447-449: “This suggests …” > This sentence should move to the Discussion chapter, since it contains the author’s interpretation of derived results. I suspect this suggestion, since the change of mangrove forest crown surface cover area showed a very significant positive correlation with urban population only in two city/county (Wenchang and Lingshui) (see Table 7).

L553-567: This paragraph describes the study limitation and future direction. It should move to the Conclusions chapter, since it is not regarded as discussion (the author’s interpretation of derived results).

L616: What is “the BigMap application”?

L622: I found many typos (distinction between upper and lower case letters, journal names mixed with full and abbreviated, etc.) in the list of references. The author should carefully read the instruction for writing and fully correct them.

Author Response

Dear Reviewer 1,

Thank you very much for your time and effort reviewing this manuscript. We have responded to the comments below. We believe the edits have resulted in an improved manuscript and our response to all the comments is provided below. In addition, we attach great importance to the advice given by reviewer. We have marked your reply with a green background color. Although we asked experts in the United States to polish the language ofthe article, some experts still raised grammatical mistakes. Therefore, we found a professional translation company to polish the article in English. See the language modification section for details. The PDF file is a detailed point-to-point response to the comments. 

Yours sincerely,

(on behalf of all co-authors)

Author Response File: Author Response.pdf

Reviewer 2 Report

Please see the attachment files.

Comments for author File: Comments.pdf

Author Response

Dear Reviewer 2,

Thank you very much for your time and effort reviewing this manuscript. We have responded to the comments below. We believe the edits have resulted in an improved manuscript and our response to all the comments is provided below. In addition, we attach great importance to the advice given by reviewer. We have marked your reply with a blue background color. Although we asked experts in the United States to polish the language ofthe article, some experts still raised grammatical mistakes. Therefore, we found a professional translation company to polish the article in English. See the language modification section for details. The PDF file is a detailed point-to-point response to the comments.

Yours sincerely,

(on behalf of all co-authors)

Author Response File: Author Response.pdf

Reviewer 3 Report

I made note of a handful of long and confusing sentences.  I think these can be addressed by splitting into two or three sentences.  The authors present percentages and measurements with a precision carried out to the 100th decimal when rounding to the nearest whole number would be appropriate.  

Line 139: I suggest using “to” instead of “-“ 108°37′ to 111°03′ E, and 18°10′ to 20°10′ N.

Line 153: I suggest the authors match the scale’s range along the X & Y axis with those mentioned on line 139. 

Line 156: I think using “Landsat time-series with a 30m spatial resolution…” simplifies this sentence

Line 157: 2^nd sentence: I recommend “All image data…” instead of “All imagery data”

Section 2.1.2

There is missing information in this section that should be included and, in my opinion, will strengthen this manuscript.  Please provide information about the date in which the images were acquired and any other pertinent scene-specific information.  Also, the authors mention acquiring imagery from adjacent years to accommodate cloud coverage but there is no information about if they were used, how they were processed and used.  This information should be included in the manuscript.

Lines 167 – 174:  This section should be addressed.  On lines 170 - 171, the authors describe classifying three bands of Landsat 5 to synthesize standard pseudocolor images.  Traditional classification of a three-band image would not produce a pseudocolor image – compositing or layer stacking the layers would yield such an image.  Also, it is unclear how these images were used in this study.

Section 2.2.2

Lines 197 – 200: This first sentence is a mixture if disassociated ideas and is difficult to understand.  I highly recommend revising or omitting.

First paragraph in this section:  I think some of this information is better suited for the introduction and the remaining removed.  I recommend beginning this section with an explanation of your deep learning model (Figure 3).

Line 234: the authors refer to ”false-color” images as “pseudocolor” in section 2.1.2.  Inconsistency in terminology should be addressed.

Section 2.2.3:

Lines 250 – 252:  It is unclear what the authors mean by  “Further, 2x3 confusion matric calculations were performed, and the precise”.  The second part of this sentence is understandable.

Section 2.2.4

I think the first sentence can be clarified by breaking it up into two sentences.  I suggest “…2015, and 2021.  The annual rate of change…”

Section 3.1.1

The classification results (mangrove areas) are too small to see in Figure 5.  This figure does not add to the discussion;  either eliminate this figure or revise it so pertinent information is visible.

Regarding Table 1, I think converting the sample counts to percentages will make it easier for readers to understand the data presented.  The years should be centered vertically – aligned with “Non-Mangrove”.  Also, I don’t think there is a need to maintain precision to the 100^th of a percent.

Section 3.1.2

Figure 6 should be revised or omitted (same comment as Figure 5).

Section 3.2.1

Are the large decreases and increases seen Table 4 due to misclassifications or are they representative of what is happening on the ground?

Section 3.2.2:

There is too much tabular information presented in Table 5.  I recommend presenting this data as a series of graphs instead.  Doing so would allow readers to easily observe the trends present.

The idea of ‘mass center offset’ is interesting, however its presentation in Figure 10 is somewhat confusing.  It is difficult to discern east/west, north/south or otherwise trends for each time period.  Would simplifying these graphs by showing the mass center offset movement based on only the first and last time periods tell the same story?

Section 4.1:

I suggest simplifying the sentence spanning from lines 475 to 481.

Line 498 – 499: I suggest the authors revise the sentence to “Therefore, we conclude that the more accurate Res-UNet provides improved mangrove forest crown surface…”

I suggest the authors round the SVM and Res-UNet values to the nearest whole number for consistency

Section 4.2

Line 535: I recommend editing to “Spatially, the changes in mangrove…” 

Lines 535 – 540:  This sentence needs to be simplified.  A few short, concise sentences are superior to one long sentence (like this one).

Line 540 – 541:  Instead of “The growth of mangrove forest crown surface coverage”, I suggest “The expansion of mangrove forest crown in each city…”

Author Response

Dear Reviewer 3,

Thank you very much for your time and effort reviewing this manuscript. We have responded to the comments below. We believe the edits have resulted in an improved manuscript and our response to all the comments is provided below. In addition, we attach great importance to the advice given by reviewer. We have marked your reply with a yellow background color. Although we asked experts in the United States to polish the language ofthe article, some experts still raised grammatical mistakes. Therefore, we found a professional translation company to polish the article in English. See the language modification section for details. The PDF file is a detailed point-to-point response to the comments. 

Yours sincerely,

(on behalf of all co-authors)

Author Response File: Author Response.pdf