Application of Image Segmentation in Surface Water Extraction of Freshwater Lakes using Radar Data
Round 1
Reviewer 1 Report
The work includes an improved approach to the analysis of satellite products. The authors set themselves a very ambitious goal. The usefulness of the test results can be large and can be used to assess the quality of the environment
Keywords:
Keywords are not adequate for the workload. I think it would be good to add "water surface detection", "water level changes"
Fig. 1. Imprecise location. There is a lack of facilities like a hydrographic network, cities. It is difficult to read where the examined object is located. More clearly indicate which objects are "river" and "lake"
Fig. 4. It is worth adding dates on the maps. My guess is that the first from the left is 2017.
2.1. Study area
There are no features of lake supply, where the fluctuations come from, what is the regime of feeding through the Gan river, climate features. After all, these elements are essential for shaping the water level of the studied lake.
3. Results and discussion
I suggest strengthening the marking of research results for implementation in nature protection. Make more sentences in a literature-based discussion.
Author Response
Point 1: Keywords are not adequate for the workload. I think it would be good to add "water surface detection", "water level changes"
Response 1: Thank you for pointing it out. I added “water area changes” to the keywords (please see line 16 - 17). Since the title already contains “surface water extraction”, I did not add the “water surface detection” according to the general rule provided by the journal that “avoid putting the same words that exist in the title”.
Point 2: Fig. 1. Imprecise location. There is a lack of facilities like a hydrographic network, cities. It is difficult to read where the examined object is located. More clearly indicate which objects are "river" and "lake".
Response 2: Sorry, we only included the 1st and 2nd order streams at the start because we did not have 4th order streams data earlier. Now we added the Jiujiang city and all the rivers connected to the lake on figure 1 and labelled all the rivers in blue marks to make them clearer.
Point 3: There are no features of lake supply, where the fluctuations come from, what is the regime of feeding through the Gan river, climate features. After all, these elements are essential for shaping the water level of the studied lake."
Response 3: We added more detailed information for the description of study area, such as temperature zone, precipitation and groundwater (please see line 118 – 121, and line 129-133).
Point 4: I suggest strengthening the marking of research results for implementation in nature protection. Make more sentences in a literature-based discussion.
Response 4: We added a paragraph to discuss how our research is associated and important to other fields in Poyang Lake (please see line 275-279). Also, we rewrote the conclusion to highlight the contributions of our study, and what other studies our method can be applied and extended to (please see line 201-293, and line 299-303).
Reviewer 2 Report
Paper "Application of Image Segmentation in Surface Water Extraction of Freshwater Lake using Radar Data" presents research in the field of surface water analysis using Otsu's method. The authors presented in the paper the real case study - Poyang Lake, the Yangtze River Basin, China.
The paper presents a very interesting research; On the other hand, the following are some facts and suggestions regarding the paper:
Suggestion 1:
For keywords, the general rule is not to put the same words that exist in the title, but rather the specific words that depict the abstract and are used in it. It is proposed that the following word be excluded: freshwater lake
Suggestion 2:
The Introduction contains details that are not usually put in that chapter. It is proposed to move the Description of the Gaussian mixture model to the chapter describing the methodological procedures.
Suggestion 3:
On the other hand, as far as the Introduction chapter is concerned, a brief description of Otsu's method is given at the end. It is proposed to add some more details of the application of Otsu's method in related scientific fields, as well as the advantages and disadvantages of the mentioned method (through literature units).
Suggestion 4:
Among other things, the paper presents the hydrological cycle of the lake. A brief description of the geological settings in the chapter "Study area" is proposed, as well as a detail on the hydraulic link of surface water and groundwater. I think that it will be much clearer and more interesting to the readers by adding these details.
Suggestion 5:
Figures does not meet high standards. As for Figure 3, it is proposed to add a description of the parameters on the coordinate axes to make it clearer. On the other hand, maps must contain coordinates. It is proposed to refine Figures 2 and 4.
Suggestion 6:
There are a large number of references in the paper that refer to the same journal - for example Remote sensing of environment journal. Please check that no more than 10 - 15 % of the references are from the same journal and no more than three references are from the same author. This is a general rule.
Suggestion 7:
In Conclusion chapter, there is no need to discuss the applied methodology, it is well known. The Conclusion would be more appropriate to state in detail the academic and scientific contribution of the paper. In the sent version of the paper, the authors are asked to clearly state which is the novelty of the paper for which the paper deserves publication.
_____ Once again praise the authors for a very interesting research.
Author Response
Point 1: For keywords, the general rule is not to put the same words that exist in the title, but rather the specific words that depict the abstract and are used in it. It is proposed that the following word be excluded: freshwater lake.
Response 1: Thank you for pointing it out. I changed “freshwater lake” into “Poyang Lake” and added “water area changes” to the keywords (please see line 16 - 17).”
Point 2: The Introduction contains details that are not usually put in that chapter. It is proposed to move the Description of the Gaussian mixture model to the chapter describing the methodological procedures.".
Response 2: We do not feel like that we should move it to the methodology, because we did not use it in our study. The Gaussian Mixture Model in the paper was to be compared with Otsu method so that we could provide a background of other conventional methods for binary classification and for water detection (Line 71 - 75),
Point 3: On the other hand, as far as the Introduction chapter is concerned, a brief description of Otsu's method is given at the end. It is proposed to add some more details of the application of Otsu's method in related scientific fields, as well as the advantages and disadvantages of the mentioned method (through literature units).
Response 3: We added sentences and references for describing more applications using Otsu method at the end of introduction. One was an application on the urban field detection (Line 94 - 98), and the other was an application on the oil spill detection over sea surface (Line 98 - 102). In the paper, we provided the advantages of using the Otsu Method as compared to the other existing methods (Line 59 - 92). Also, we discussed about the disadvantage of applying the Otsu Method (Line 92 - 94). However, those advantages and disadvantages were not fully and clearly discussed in remote sensing filed yet. I think this could be the one of the contributions of our study.
Point 4: Among other things, the paper presents the hydrological cycle of the lake. A brief description of the geological settings in the chapter "Study area" is proposed, as well as a detail on the hydraulic link of surface water and groundwater. I think that it will be much clearer and more interesting to the readers by adding these details.
Response 4: We added more hydraulic information for the description of study area, such as temperature zone, precipitation and groundwater (please see line 118 – 121, and line 129-133).
Point 5: Figures does not meet high standards. As for Figure 3, it is proposed to add a description of the parameters on the coordinate axes to make it clearer. On the other hand, maps must contain coordinates. It is proposed to refine Figures 2 and 4.
Response 5: We added detailed descriptions both in paragraph (please see line 260 - 261) and figure 3 to explain the axis and histogram. We added coordinates for figure 4. However, I did not change figure 2, because we described the coordinate of each point in the caption. Since I zoom in the figure in order to clearly show each point, the difference of latitude/longitude across each sub-figure is very small, it seems redundant and not helpful to add coordinate axes on the figure frame.
Point 6: There are a large number of references in the paper that refer to the same journal - for example Remote sensing of environment journal. Please check that no more than 10 - 15 % of the references are from the same journal and no more than three references are from the same author. This is a general rule.
Response 6: Thank you very much for this suggestion. We removed and replaced all the journals that beyond 10 % of the total references. Currently, we have 51 references in total and the most one is Remote Sensing which occurred 5 times. And we double checked there were no more than 3 references from the same author.
Point 7: In Conclusion chapter, there is no need to discuss the applied methodology, it is well known. The Conclusion would be more appropriate to state in detail the academic and scientific contribution of the paper. In the sent version of the paper, the authors are asked to clearly state which is the novelty of the paper for which the paper deserves publication.
Response 7: We added a paragraph to discuss how our research is associated and important to other fields in Poyang Lake (please see line 275-279). Also, we rewrote the conclusion to highlight the contributions of our study, and what other studies our method can be applied and extended to (please see line 201-293, and line 299-303). We deleted some detailed sentences for the method according to your comments as well.
Overall, your review is super helpful and important for us, thanks a lot for your time and suggestions again.
Reviewer 3 Report
I enjoyed reading the paper by Zhou et al, which attempted to analyze the advantages and disadvantages of the Otsu segmentation algorithm. It is of interest for the community and I recommend it for publication after considering the following:.
To help place the findings into a broader international context (in the Introduction section) can the authors provide an objective-focused tabular literature review of relevant research: including the current study as the last row in this proposed table. Doing this should make it more apparent to readers (and reviewers) what the new contribution of this study.
For examples of objective-focused tabular literature reviews, please see Table 1 of the following papers.
This reviewer found the selection of the optimal threshold for 2020 based on histogram is not enough robust. Can the others confirm this threshold with other analyses?
Can the authors validate the outputs of figure 4?
Is this Otsu segmentation algorithm transferable? Can you try it over other regions than Poyang Lake?
Author Response
Point 1: To help place the findings into a broader international context (in the Introduction section) can the authors provide an objective-focused tabular literature review of relevant research: including the current study as the last row in this proposed table. Doing this should make it more apparent to readers (and reviewers) what the new contribution of this study.
For examples of objective-focused tabular literature reviews, please see Table 1 of the following papers.
Response 1: Based on your suggestions, we added sentences and references for describing more applications using Otsu method at the end of introduction. One was an application on the urban field detection (please see line 94 - 98), and the other was an application on the oil spill detection over sea surface (please see line 98 - 102). In the paper, we provided the advantages of using the Otsu Method as compared to the other existing methods (please see line 59 - 92). Also, we discussed about the disadvantage of applying the Otsu Method (please see line 92 - 94). However, those advantages and disadvantages were not fully and clearly discussed in remote sensing filed yet. This was the one of the contributions of our study.
Point 2: This reviewer found the selection of the optimal threshold for 2020 based on histogram is not enough robust. Can the others confirm this threshold with other analyses?
Response 2: The Otsu Method is an image segmentation method, which has been widely applied in the Computer Vision field for binary classification. Due to its well-known correctness in algorithm, we applied it over the radar data for differentiating water from non-water pixels. In the paper, we provided a detailed mathematical and statistical proof for the correctness of the Otsu Method (please see line 155 - 222), which can theoretically find a global optimal threshold based on the density histogram. Thus, it is sufficient to confirm that the threshold provided in the paper, which is derived from the Otsu Method, is a valid global optimal threshold.
Point 3: Can the authors validate the outputs of figure 4?
Response 3: We used the ground truth data to validate the classification results (please see line 211-222) and confusion matrix to assess the accuracy (please see line 224-258). If the question means using other research or other methods, the answer would be no. First, the most updated publication about surface water extraction of Poyang Lake only dated to year of 2016. Second, we did not run comparative analysis with other methods because we listed this purpose in the future work (please see line 294 - 303). However, anyone is encouraged to use our open and free code to reproduce our work (please see line 203), which is one of our main contributions -providing reproducible and transparent results- in this study. On the other hand, previous research seldom has provided reusable algorithm and scripts as to our knowledge.
Point 4: Is this Otsu segmentation algorithm transferable? Can you try it over other regions than Poyang Lake?
Response 4: As we mentioned in the Response 2, the Otsu Method is a statistical algorithm referred from the Computer Vision fields, so the Otsu Method is theoretically transferable to not only other regions than the Poyang Lake but also any image processing fields involving binary classification. The mathematical and statistical proof in the paper is a general correctness proof (please see line 155 - 222), which is independent of a specific density histogram or distribution. Thus, such proof guarantees for the generality of the performance of the Otsu Method over other regions. However, the exact performance of the Otsu Method over other regions may depend on the quality of the input data and the constrains we mentioned in this paper (please see line 57-58), such as noise in the input density histogram. Due to the fact that we focused more on the surface water of the Poyang Lake and on the purpose of the conservation of the Poyang Lake, it is less necessary for us to carry out more experiments over other regions in this paper. However, we encourage other readers or reviewers to try to apply the Otsu Method for surface water detection in other regions, as the Otsu Method is a theoretically general image segmentation method.
Round 2
Reviewer 2 Report
Based on the received report and revised version of the paper, I believe the paper is appropriate for publication.
However, I suggest some minor improvements before accepting.
If the autors mentioned not using the Gaussian mixture model, the equations don’t attach as much importance then. It is proposed to be excluded from the Introduction.
Author Response
Point: If the autors mentioned not using the Gaussian mixture model, the equations don’t attach as much importance then. It is proposed to be excluded from the Introduction.
Response: Thank you for your suggestion again, we have removed the equations from the introduction, and replaced with specific descriptive sentences for the method (please see line 70 - 77).
Reviewer 3 Report
The authors addressed all my comments.
Author Response
Point: English language and style are fine/minor spell check required.
Response: Thank you for your suggestion, a native speaker has checked the whole document and revised words, sentences and grammar. We have highlighted all the changes in color.
