Monitoring and Forecasting of Urban Expansion Using Machine Learning-Based Techniques and Remotely Sensed Data: A Case Study of Gharbia Governorate, Egypt

Round 1

Reviewer 1 Report

In general, I think that this study presents a useful step forward for the community and that it builds onto the present state of knowledge. I have a couple of minor concerns with the presented introduction and methodology that should be addressed. My minor comments and questions are as follows:

• What kind of influencing factors were available of urban expansion over the study area. How can you utilize, and characterize your methodology?
• Why did you use Landsat images to analyze historical LULC change? Very high spatial resolution satellite sensors (e.g WV02 imagery) provide opportunities to observe multiple spatial scales and temporal frequencies for remote sensing applications. You can write about it introduction section

Zhang, et al. 2020: Epstein, H.E.; Jones, B.M.; Jorgenson, M.T.; Kent, K. Transferability of the Deep Learning Mask R-CNN Model for Automated Mapping of Ice-Wedge Polygons in High-Resolution Satellite and UAV Images. Remote Sens. 2020, 12, 1085.

Ehsan, et al. 2020. Use of Very High Spatial Resolution Commercial Satellite Imagery and Deep Learning to Automatically Map Ice-Wedge Polygons across Tundra Vegetation Types. J. Imaging 2020, 6, 137.

 

• The Influence of fusion on classification (e.g.: vegetation/tundra ) problem for image application you can discuss in the introduction section. You can see the latest high impactful research to see the Influence of fusion on classification for image application.

 

Witharana, Chandi, et al. "Understanding the synergies of deep learning and data fusion of multispectral and panchromatic high resolution commercial satellite imagery for automated ice-wedge polygon detection." ISPRS Journal of Photogrammetry and Remote Sensing 170 (2020): 174-191.

 

Yang, J.; Zhao, Y.-Q.; Chan, J.C.-W. Hyperspectral and Multispectral Image Fusion via Deep Two-Branches Convolutional Neural Network. Remote Sens. 2018, 10, 800. https://doi.org/10.3390/rs10050800

Deur, M.; Gašparović, M.; Balenović, I. An Evaluation of Pixel- and Object-Based Tree Species Classification in Mixed Deciduous Forests Using Pansharpened Very High Spatial Resolution Satellite Imagery. Remote Sens. 2021, 13, 1868. https://doi.org/10.3390/rs13101868

 

• Can you explain more about overfitting issues?
• Can you please provide a table for validation/training/testing samples along with other information such as sample source, training and validation sites, repository, etc?

 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors of this manuscript used the LCM in Terrset to predict the land cover/use change in the area of Gharbia Governorate in 2033 and 2048. In methodology, there is no novelty, because the classification method and the model for prediction have existed for a long time and they used a software to conduct the prediction. However, as an application study, it may have some value, such as providing related planners some reference data and insights about the possible future change trends of the land cover/use in the study area with different scenarios. So whether this paper is suitable for publication, it depends on how the authors can highlight the importance of their findings.

Some main concerns are needed to be solved to make the article more reasonable. 

1. What is the original contribution of this paper? The classification is standard, and model applied is commonplace. How does this work advance the science?
2. Why do you choose the year of 1991, 2003, and 2018, and predict for 2033 and 2048. Give some practical reasons to make the time points reasonable.
3. The authors mentioned before that it is important to make a study here because of agriculture. But you didn't classified agriculture land separately. How can you evaluate the results for agriculture development? 
4. The length of the manuscript is too long, almost double size of a common paper. The authors provided too many basic information and makes the key contents not obvious and hard to catch. Besides, you reviewed a lot of related researches but fail to show your advance compared to these studies. Anyway, I strongly suggest the authors to shorten your manuscript to make your main points clearer.
5. I don't agree with the method the authors used for model validation. Please refer to the paper: "Death to Kappa: Birth of quantity disagreement and allocation disagreement for accuracy assessment". The validation method is not vary acceptable now to use Kappa coefficient.
6. The Fuzzy TOPSIS Analysis: t's totally another analysis. No connection with the previous analysis in methodology. And the results are not detailed (only district level). I don't think it's needed here.
7. Figure 4d and 6d are not necessary. The relative tables have already show the more detailed information.
8. In discussion, the authors mentioned "The Jaccard similarity coefficient was 52%. The Jaccard coefficient statistic much better estimates the fit between the ob served and simulated LULC maps. The usually acceptable values of Jaccard coefficient are > 60%. However, the results were considered adequately acceptable and the model could be adopted for future simulation." I'm very confused here because I didn't find any explainations here. Why is it acceptable even if it is below the normal level?
9. 3.4. Linear Regression Analysis: this section is confused. Is this the change of vegetation? Then it's wrong to say "agriculture land change" here. In your classification, agriculture land has not been classified as a single category. How can you get "agriculture land change"?
10. The Urbanization Risk Map is actually based on Fuzzy TOPSIS. That means it is totally based on subjective opinions without reference to the available natural and social data. In this case, I'm not confident with the result and I suggest to combine with some existed data from government.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors have made obvious improvements based on the previous comments and now the paper is significantly improved, especially for emphysizing that Fuzzy TOPSIS Analysis can evaluate the districts lack of satellite data. I still have two concerns for the manuscript which I think have not been well explained in this round.

1. The reason for choosing the year of 1991, 2003, and 2018: I understand it is because of data availability and I noticed you only selected the images in June. I think this point should be mentioned so that the readers can accept considering you want to use images in the same month. Also, why the images of June is suitable for the analysis, this is needed to be explained.
2. The categories of the classification: I haven't been convinced why you didn't separate the category of vegetation into more detailed categories including agriculture land, grassland, and forest. As I mentioned in the previous round, you emphysized the importance of agriculture here. Then I think you need to classify it as a single category to carefully check it. Althogh you made a explaination, vegetation doesn't equal to agriculture. So I'm still not satisfied with this point.

Author Response

Please see the attachment

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

Authors are presenting and interesting research and case study.

After reading the paper with great interest, we can testify that the ideas' flow is clear and logic, the structure of the paper is following the standards, and the analysis part is sufficient. However, from our point of view, and for the betterment of the research paper, authors are highly invited to take into consideration the language level quick revision, a revision of the introduction (please do provide sufficient background), enrich the analysis and conclusion parts. Finally, some relevant references (actual and basic) on ML are missing.

In addition, a profound revision of some minor errors are highly recquired:

For example, in Eq. (1), respect the parameters' format (same as in text)... k is not K, (same for other parameters Po, Pe)...

There are some numbering confusion at Figures level:

For example:

Figure 6 caption is under other figure, right! Figure 6a and Figure 6d are not cited in the text.

idem for Figure 7 and others...

 

 

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

I think the authors carried out a large amount of work to predict Urban Expansion with machine learning over the Gharbia Governorate, Egypt. The paper is generally well-written in an understandable way. I have included several suggestions and recommend the manuscript for publication after the following major changes:

• The idea of the manuscript is not new, which is similar to other studies. What will be the impact of these evaluation results for the data-gap regions? Also please introduce other Urban Expansion-related research work for other study regions which will provide more insight which will help to explain the novelty of this work or improvement with respect to the previously derived methodology?

Sankhala, et al. 2014: Evaluation of urban sprawl and land use land cover change using remote sensing and GIS techniques: A case study of Jaipur City, India. J . Int. J. Emerg. Technol. Adv. Eng. 2014, 4, 66–72

Jahan et al. 2021: Surface Runoff Responses to Suburban Growth: An Int Shrestha, S.; Cui, S.; Xu, L.; Wang, L.; Manandhar, B.; Ding, S. Impact of Land Use Change Due to Urbanisation on Surface Runoff Using GIS-Based SCS–CN Method: A Case Study of Xiamen City, China. Land 2021, 10, 839egration of Remote Sensing, GIS, and Curve Number. Land 2021, 10, 452

• The authors have not provided sufficient motivation for why Fuzzy TOPSIS, SVM were needed. What is the uniqueness of the techniques and their potential impacts, over other established state-of-the-art (e.g. machine and deep learning algorithm-based)? The authors should explain with a couple of new paragraphs on this aspect in the introduction section. Also, you need to provide more literature reviews in terms of research gap/limitation.

 

• Your study area has a high dependency on local climate, topographic complexity. Can you provide detail climatic information for the selected study areas? You should show Köppen–Geiger climatic zones on the map.

 

• Is there any significant error due to variability and uncertainty introduced by orographic effects?

 

• Can you provide a high impactful schematic diagram for the Fuzzy TOPSIS method to understand the algorithm where the big impact of the results can be presented?

 

• can you please explain more about the uncertainty of your ML methods?

 

• In figure 2 you introduced RGB band combinations, but you didn’t discuss them. There are lots of practical challenges that apply to three-band combinations in imagery study. A recent study highlighted the importance of the band combinations in the use of multispectral datasets on machine learning model prediction accuracy for remote sensing applications. Multispectral satellite imagery with three channels, preferably true-color (red, green, and blue) or false-color (green, red, near-infrared) composite with the radiometric depth of 8 bit or 16 bit have a great impact in ML-based techniques. It can be R,G,B or G,R,NIR but you dint introduce any recent studies. You can add one paragraph and explain about high complexity of these models with more references?

Abdalla, Alwaseela, et al. "Color Calibration of Proximal Sensing RGB Images of Oilseed Rape Canopy via Deep Learning Combined with K-Means Algorithm." Remote Sensing 11.24 (2019): 3001.

Park, Ji Hyun, et al. "RGB Image Prioritization Using Convolutional Neural Network on a Microprocessor for Nanosatellites." Remote Sensing 12.23 (2020): 3941.

Bhuiyan, et al.. "Understanding the Effects of Optimal Combination of Spectral Bands on Deep Learning Model Predictions: A Case Study Based on Permafrost Tundra Landform Mapping Using High Resolution Multispectral Satellite Imagery" Journal of Imaging 6, no. 9: 97, 2020.

 

 

• Also, multispectral images are always partitioned into regular patches with smaller sizes and then individually fed into ML for semantic segmentation. If these images are independent of one another in terms of geographic spatial information, what would be the challenge and recommendation?
• Why did you select Landsat image instead of high-resolution satellite imagery? What is the Spatial resolution of your Landsat Imagery?
• How did you correct your images were geometrically and atmospherically? can you explain in detail?
• can you justify with more references, You mentioned: “SVM which is one of the most common machine learning image classifiers”?

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Dear authors, I have read your artilcle and it is well written and presented. The results are valid and make sense. At this stage I think it is ready for publication with any major minor changes. 

 

Author Response

Thanks, dear reviewer for your evaluation. I appreciate that. Especially the authors did their best to produce this paper which was revised well and carefully. 

Round 2

Reviewer 2 Report

The authors significantly improved the quality of the paper by addressing most of the previous comments. This research work will be very effective for the remote sensing community. I recommend the manuscript for publication after minor changes.

• Recent few studies also investigated LULC related to your studies? You can add after line 55.

Imran, H. M., et al. "Impact of Land Cover Changes on Land Surface Temperature and Human Thermal Comfort in Dhaka City of Bangladesh." Earth Systems and Environment (2021): 1-27.

Lopez-Saez, J.; Corona, C.; Eckert, N.; Stoffel, M.; Bourrier, F.; Berger, F. Impacts of land-use and land-cover changes on rockfall propagation: Insights from the Grenoble conurbation. Sci. Total Environ. 2016, 547, 345–355.

• Can you provide high-resolution figure 9 with the correct font?
• In the discussion section can you add specific limitations of your methods with strong recommendations.
• You need to check your reference format very carefully. Please correct Reference 1,2,3,19 according to the journal format.

Ref 9 also showing wrong. It would be:

“9. Jahan et al. :Surface Runoff Responses to Suburban Growth: An Integration of Remote Sensing, GIS, and Curve Number. Land 2021, 10, 452.”

Author Response

Please see the attachment

Author Response File: Author Response.pdf