Identifying Urban and Socio-Environmental Patterns of Brazilian Amazonian Cities by Remote Sensing and Machine Learning

Round 1

Reviewer 1 Report

The ecological environment protection in tropical rainforest areas has significant global ecological significance. The Brazilian Amazon Basin holds a exemplary representative position in the global tropical rainforest environment.  I am honored to see the important research results that the authors have brought to the academic community. Obviously, the authors have expressed the latest scientific understanding through careful consideration and technological application.  I think this article responds well to the concerns that scholars from other regions around the world may have about the Amazon region. Propose the following questions for further reference by the authors.

1-The selection criteria for two types of cities will have an impact on future research. Therefore, the author needs to explain to the readers the reasons for your choice.

2- Classification standards can appropriately increase the judgment process of researchers, especially how expert opinions and machine learning classification are combined, which is worth further elaboration.

3- There is currently a possibility of further reduction in the length of the article. The author can try to discuss in a more concise way.

Author Response

See letter attached.

Reviewer 2 Report

Abstract needs to be rewritten.

Contributions should be added point-wise. Organization of paper can be added.

Related work can be enhanced. It is specifically focused on only 5 to 6 methods. It should be comprehensive/summary of recent works in this area.

Figure 1 shows map which looks like taken from Google Maps or relevant sites. It should be cited.

Methodology is sufficiently explained.

Results and discussion are well presented and explained. Comparison with SOTA methods should be done.

Conclusion can be minimized and summarized precisely. Future work can be added.

Moderate English editing is required. 

Author Response

See letter attached.

Author Response File: Author Response.docx

Reviewer 3 Report

I found this an intriguing case study that should be of interest to both GIS/Remote sensing and Critical GIS audiences.  The application of GEOBIA-based methods together with field work provides important data.  I was fascinated by the 'social construction of science' undercurrents, but became a bit muddled trying to follow the technical details of your machine learning methods.  The high level model is clearly presented (and even better, woven seamlessly into social commentary), and there are plenty of small details regarding exact parameters used, but there seems to be a missing middle ground.  Could you more clearly chart out the GIS workflow, step to step, indicating which programs and tools were used with which datasets?  I also think the manuscript would benefit from more direct discussion of the various spatial resolutions at play, and how they were resolved.  I couldn't always tell, for example, what cell size was utilized in the different raster grids.  My last general comment is grammar:  while the manuscript reads well overall, please proof one more time for a handful of grammatical errors that still remain.

I also collected a list of specific comments that I include below:

·      Background section, lines 77-130.  Expand (slightly) your literature review of land use classification with pixel and object-based approaches, beyond Brazil.  Ellis and Ramkutty (2008) (for their Anthromes dataset) and Blaschke et al. (2014) (for the GEOBIA concept) come to mind.

-  Lines 195-206. The justification for treating environmental and urban morphology variables separately needs to be clarified.

·      Assessment Criteria (following line 328) --- clarify where  you get the data?  (especially the inadequate water/garbage data)

·      3.3 Creating and Integrating variables into the cell grid:  what was the cell size?

·      Methods represented in figure 4 (and their descriptions) – not always clear which data sources you are using for which assessments

·      You lost me in lines 344 to 357.  Was this in QGIS? Python? Perhaps provide more details about specific commands used?  Didn’t follow what was vector or raster, either.

·      Line 436 -- The “base of Subnormal Agglomerations” – is this a dataset? 

•  
• What does it mean to “apply Adaptive Axial Maps” as an assessment method (about page 9)? 
•  
• Figure 6, maps c) and d) –one of them must be mislabeled, or caption is wrong, or both? 
• Lines 470-476:  I’m not familiar with k-modes algorithm, so I'm having a hard time imagining what “experimentation” entails.  What changed between each iteration? 
•  

Generally the English is fine, save for a few grammatical errors.

Author Response

See letter attached.

Author Response File: Author Response.docx

Reviewer 4 Report

Well articulated paper. The objectives of the study could be listed in the last paragraph of the introduction. Mention why unsupervised classification techniques were used.

No major typo errors notes, only minor ones. Language is clear and free flowing.

Author Response

See letter attached.

Author Response File: Author Response.docx