Land-Cover-Change Detection with Aerial Orthoimagery Using SegNet-Based Semantic Segmentation in Namyangju City, South Korea

Round 1

Reviewer 1 Report

The manuscript is good, the authors evaluate the  Land Cover Change Detection with Aerial Orthoimagery Using SegNet Based Semantic Segmentation in Namyangju City, South Korea. It is an interesting and great contribution to the scientific community; however, the introduction, methodology and references of the paper should be improved. Still there are many issues present in the manuscript which should be explained properly. The manuscript needs some minor revisions as given below:

 

·         The text of this paper in general needs a thorough review, as there are multiple spelling and grammatical errors. Many sentences do not mean any sense. Moreover, there are several sloppy errors that should be fixed.

·         Line 17-18; It is not necessary, it is presented in first para. So please delete this line.

·         Line 20-21; Confusing, Rephase this sentence.

·         Please explain Land Use in introduction.

·         More research background and motivation should be added to the Introduction section. Although, I propose some new papers must be added in the reference list and text which will also help you to make it more intriguing such

https://doi.org/10.3390/land11050595, https://doi.org/10.1016/j.pce.2022.103117.

·         Line 74-84; Move this paragraph in method section.

·         Why are you conduct this study in Namyangju City, Please explain in your study area background in introduction section.

·         Line 112; Breiman (2001), set this as journal style

·         Line 122; Chervonenkis and Vapnik (1971) very old reference, please add updated reference

·         Table 3; Why are you select 2 years 2010 and 2012 ?? 1^st you selected pervious year data, you must select least year for scientific results. It is hard to understand the land cover change in 2 year gap. I suggest please select the other year data for best results like you selected such years 2000 and 2020 and then check LULC changes.

·         In results and discussion section, your citation is not correct, please se carefully and also set as journal style.

·         In discussion section; Discussion: As per the instruction given by the journal “The findings and their implications should be discussed in the broadest context possible and the limitations of the work highlighted”

·         The conclusion is too general. What are the key findings of this study?

·         Write the future recommendation of your research in conclusion.

 

Overall, the study conducted is interesting but a minor revision of the entire manuscript is essentially required for publication in this journal. Hence, I recommend reconsideration after a minor revision of the manuscript.  

 

 

Author Response

Response for reviewers #1

 

Thank you for a detailed and thorough review for paper.

I tried to make an effort to mirror the reviewer’s comments on the paper and two times of English language and style correcting were performed by requesting a professional institution.

First, spelling and notation error, comments on expression in English and references error were reflected according to the reviewer’s comments. I will try to reduce these kinds of error by reviewing paper repeatedly. Besides, the contents of paper were corrected to reflect the reviewer’s comments.

 

 

1. The text of this paper in general needs a thorough review, as there are multiple spelling and grammatical errors. Many sentences do not mean any sense. Moreover, there are several sloppy errors that should be fixed.

 

• Spelling and comments on expression in English and references error were reflected according to your comments. And we performed a "spelling and grammar" check for minor errors.

 

 

 

2. Line 17-18; It is not necessary, it is presented in first para. So please delete this line.

 

• As you commented, we deleted unnecessary information (The objective of this study was to select the model with the highest accuracy to efficiently detect land cover changes.) included in Abstract.

 

 

 

3. Line 20-21; Confusing, Rephase this sentence.

 

• We modified sentence of Line 19-20 to “The study areas were the Namhan and Bukhan river basins, where significant urbanization occurred between 2010 and 2012” in Abstract.

 

 

 

4. Please explain Land Use in introduction.

 

• In general, land use and land cover are used with the same meaning, and in many previous literatures, they are often used as land use land cover (LULC). Therefore, we modified the “land cover” of line 36 and 39 to LULC.

 

 

 

5. More research background and motivation should be added to the Introduction section. Although, I propose some new papers must be added in the reference list and text which will also help you to make it more intriguing such

https://doi.org/10.3390/land11050595, https://doi.org/10.1016/j.pce.2022.103117.

 

• As you commented, we added related research background and motivation you recommended papers in line 45-48 of Introduction (LULC research plays an important role in meteorology and the sustainable management and exploitation of natural resources. Changes in LULC must be analyzed to manage and develop a sustainable ecosystem, and understand the social, economic, and environmental consequences of the changes) and reference lists.

 

 

 

6. Line 74-84; Move this paragraph in method section.

 

• According to your review, we moved Line 82-89 to Methodology section in modified manuscript (SegNet, RF, and SVM were trained using aerial orthoimagery and a level-3 LULC map of the Namhan and Bukhan river basins. Aerial orthoimagery obtained in 2010 and 2012 was used for land cover classification and change detection. The training and validation datasets for each model included the river basins, but excluded the area around Namyangju city, which was used as a test dataset. Parameter optimization was used to select hyperparameters for each model, and land cover was classified as urban, crops, forest, or water. The best-performing of the three models was used to classify land cover and detect changes around Namyangju city.).

 

 

 

7. Why are you conduct this study in Namyangju City, Please explain in your study area background in introduction section.

 

• According to your review, we modified the reason for choosing Namyangju city as the study area to “Namyangju city is in the West of the study region, where many crops and forest areas have been transformed during rapid urbanization and deforestation occurring between 2010 and 2012. Land cover change detection techniques were tested around Namyangju city” in lines 135-138 of 2. Study area and datasets

 

 

 

8. Line 112; Breiman (2001), set this as journal style

 

• We confirmed the reference style of sustainability and we modified reference style (Breiman proposed an RF that comprised an ensemble model with a tree-type classifier) in lines 112-113 of Methodology.

 

 

9. Line 122; Chervonenkis and Vapnik (1971) very old reference, please add updated reference

 

• We confirmed reference we used, we modified the sentence (SVMs were introduced by Vapnik et al. [32], and perform well in classification and regression tests) in lines 122-123 of Methodology and updated the reference to “Vapnik, V.; Guyon, I.; Hastie, T. Support vector machines, Learn, 1995, 20(3), 273-297.” in reference lists.

 

 

10. Table 3; Why are you select 2 years 2010 and 2012 ?? 1^styou selected pervious year data, you must select least year for scientific results. It is hard to understand the land cover change in 2 year gap. I suggest please select the other year data for best results like you selected such years 2000 and 2020 and then check LULC changes.

 

• There are two main reasons why we selected land cover changes in 2010 and 2012.

1. The Namhan and Bukhan river basins, especially Namyangju city, have changed large areas of forest destruction and urbanization have occurred due to rapid urbanization since 21^st
2. The level-3 LULC map for the Namhan and Bukhan river basins provided by the Ministry of Environment were produced and provided using satellite images and aerial orthoimagery in 2010 and 2012, respectively. However, because it has not been updated since, a good quality reference dataset could not be obtained.

 

 

 

11. In results and discussion section, your citation is not correct, please se carefully and also set as journal style.

 

• As you mentioned earlier, we confirmed all the references cited in the manuscript and those are modified to be set as journal style.

 

 

 

12.  In discussion section; Discussion: As per the instruction given by the journal “The findings and their implications should be discussed in the broadest context possible and the limitations of the work highlighted”

 

• We modified much of overall structures and contents of previous manuscript and we request you to review new manuscript again (Urban infrastructure has increased, and forests have decreased, owing to urbanization in the Namhan and Bukhan river basins, as revealed by the changes in LULC revealed in this study. It may be possible to preserve the ecosystem and restore biodiversity through efficient urban development and forest preservation.) in lines 355-358 of Discussion.

 

 

13. The conclusion is too general. What are the key findings of this study?

 

• As you commented, we added the key findings of our study in lines 373-375 of conclusion (Based on these results, the South Korea land cover map can be updated every 2 years, and high-resolution land cover maps can be produced using not only aerial orthoimagery, but also satellite data.).

 

 

 

14. Write the future recommendation of your research in conclusion.

 

• As you commented, we added our future recommendation in lines 375-376 of conclusion (This study could inform guidelines for urban management systems and the sustainable development of forest and crops in South Korea.).

 

 

Reviewer 2 Report

The goal of the research is clear, but the applied tools, and that is why the results not accurate.

The authors mention a serious limitation of the study: there are other models which accuracy is better (see the last paragraph before the Conclusions) than SegNet. In this context, the objectives of the paper are not clear. 

Apart from this, the comparison of the 3 models is well done, but its benefit is not clear. 

Author Response

Response for reviewers #2

 

Thank you for a detailed and thorough review for paper.

I tried to make an effort to mirror the reviewer’s comments on the paper and two times of English language and style correcting were performed by requesting a professional institution.

First, spelling and notation error, comments on expression in English and references error were reflected according to the reviewer’s comments. I will try to reduce these kinds of error by reviewing paper repeatedly. Besides, the contents of paper were corrected to reflect the reviewer’s comments.

 

 

1. The goal of the research is clear, but the applied tools, and that is why the results not accurate.

 

• In this study, SegNet, a deep learning technique, and RF and SVM, a machine learning technique, were selected for land cover classification and change detection. Although there are differences between deep learning and machine learning techniques, various studies in Discussion have produced various land cover maps using machine learning or deep learning models. The results of these studies showed that the deep learning model was higher than the machine learning model, and the results of this study were also same to those of previous studies.

 

 

2. The authors mention a serious limitation of the study: there are other models which accuracy is better (see the last paragraph before the Conclusions) than SegNet. In this context, the objectives of the paper are not clear.

 

• As you commented, we deleted unnecessary sentences included in Discussion (This study had some limitations. Land cover classification and change detection was performed using SegNet, a semantic segmentation technique proposed in 2017. According to Lee and Lee (2020) and Rousset et al. (2021), higher land cover classification accuracy can be obtained by the UNet and DeepLab V3+ semantic segmentation models than by SegNet [47, 48]. Additionally, better results can be achieved if land cover classification and change detection analysis are performed using high-performance models.) and reference lists (Lee, S. H.; Lee, M. J. A study on deep learning optimization by land cover classification item using satellite imagery. Korean Journal of Remote Sensing, 2020, 36(6_2), 1591-1604.; Rousset, G.; Despinoy, M.; Schindler, K.; Mangeas, M. Assessment of deep learning techniques for land use land cover classification in southern new Caledonia. Remote Sensing, 2021, 13(12), 2257.)

 

 

 

3. Apart from this, the comparison of the 3 models is well done, but its benefit is not clear.

 

• As you commented, we added the advantages of the three models in the field of land cover classification to the introduction (Specifically, SVM and RF are widely used for land cover classification (lines 63-64 of Introduction); According to Garcia-Garcia et al., SegNet is widely used in the field of image recognition and classification because it is more accurate and efficient than most semantic segmentation models (lines 73-75 of Introduction) and reference lists (Thanh Noi, P.; Kappas, M. Comparison of random forest, k-nearest neighbor, and support vector machine classifiers for land cover classification using Sentinel-2 imagery. Sensors, 2017, 18(1), 18.).

Round 2

Reviewer 2 Report

The revised version became much better in many aspects. I hope that you can feel it too.