Combining Remote-Sensing-Derived Data and Historical Maps for Long-Term Back-Casting of Urban Extents

Round 1

Reviewer 1 Report

This is an article submitted to the section of urban remote sensing. This is a meaningful and useful research by integrating remote sensing data and historical maps. The purpose of this article is to propose a framework to further automatically generate historical urban extents for the early 20^th century. The followings are the unknown or uncertainties for this article, and it is hard to confirm the results are accurate or not.

1. Line 100. It would be great to illustrate “back-casting strategy.” What are the differences from current methods? What are the benefits of applying “back-casting strategy”?
2. Line 101. What does it mean that “built-up areas derived from the GHSL in 1975?” The results of urban extents for the early 20^th century is based upon GHSL 1975?
3. Line 104. What are the differences of evaluating “urban growth” and “urban shrinkage”? The results indeed might be bias for different algorithm.
4. Line 105 and line 120. Why the extraction of historical urban areas is based upon 100 x 100 tiles? So, GHSL spatial unit is 30 x 30 m? What are the calibration process?
5. Line 120-121. The article applies present or historical maps (1975, 1990, 2000, and 2014), and why? Figure 9, the urban areas is a continuous results which is based upon the four years’ remote sensing?
6. Line 115. Maybe authors could bring up the framework to increase readability for there are so many dataset (based upon different years, different scales, …) applied in this study.
7. Line 123-124. Besides the four continent, why select these 6 study areas?
8. Figure 1 and Figure 2. The historical maps are quite different in these six study areas. What are the common maps and what are particular maps?
9. Line 181. What are the data sources of HYDE database? It seems that HYDE databased could date back 2010 to 10,000 BC. Why not the study applies HYDE for both US and non-US study areas?
10. Line 194. It would be better if authors could provide structural flow chart for the preprocessing.
11. Line 225-226. What is the “simple decision rule?” Are there any other common rules is doing similar determination?
12. Figure 4. It is hard to catch how author came up the “build-up in 1900”?
13. Line 259 and 279. Is there any calibration process?
14. Line 283. What is ROC analysis? Why apply clustering analysis?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The topic is very interesting and relevant in the light of urban growth and its consequences. the article is well-written, all its parts provide the reader with adequate description of applied methods, datasets and obtained results.

Did authors detect any pattern in the used historical maps that could affect their use in the future urban areas detection using the proposed method? ... such as map scale, map colours or cartographic textures of urban areas. It would be very useful if they could formulate recommendations or point out issues to be aware of.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Interesting publication worth being published! Considering the focus lying on spatial processing of thematic maps I wonder whether the article would better fit to a Geomatics Journal.

The paper is generally well structured and good to read. In some parts, however, some issues should be adressed :

• The Introductions provides all relevant information for a research paper, but it could be better structured.
  1. It would be helpful to give some more information regarding the city planner’s perspective. What are requirements when applying back-casting methods? Area-related information is not at the first place whereas functional properties of city elements usually play a bigger role. Having an idea about information requirements is also crucial for identifying research gaps as a precondition for this research
  2. In research papers one would expect clearly defined research questions.
  3. The final paragraphs in the Intro read like belonging to the Methods chapter/Results chapter respectively
• For better understanding of the methods applied I would prefer having data flow or process diagram
• A critical discussion of the framework and the results when using it is missing or somehow hidden in the Conclusions. Pros and cons of algorithms and procedures applied discussed in the light of other concepts published in this domain will be helpful. How robust a algorithms when quality og maps changes etc.? Again, one could ask whether it makes sense having urban extents available without getting relevant functional data about city elements in the area

Author Response

Please see the attachment.

Author Response File: Author Response.pdf