A New Graph-Based Fractality Index to Characterize Complexity of Urban Form

Round 1

Reviewer 1 Report

A new graph-based fractality index to characterize complexity index to characterize complexity of urban form

This professionally written paper presents an enhanced methodology for describing and evaluating the complexity of urban morphology. The idea is to use indicators that have been proposed so far for that purpose, especially fractal dimensions, and introduce a spatial dimension by combining fractals and graph theory. The building footprints of Greater London are used for a through illustration of the methodology, and deep learning with CNNs drives the analysis. On page 2 the contributions of the study are correctly summarized.

Despite the interest of the methodology per se, a thought lingers that I was not able to resolve. So what? What are the insights about human behaviour in cities or planning to be gleaned from the analysis? What are the practical implications of a complex urban morphology pattern vs. a simpler one? How can building footprints by themselves have a dynamic dimension (e.g. rates of growth)? What about the fact that building footprints are very insufficient for describing the real, three-dimensional city? What other areas of application of the methodology present themselves? And so on. The authors attempt to address the ‘so what’ question in the last paragraph of the Discussion, and mention in passing potential areas of application ranging from urban quality of life and transportation issues, to climate and tumors. Perhaps, but I’m not convinced.

In summary, despite the above misgivings, I think that the paper could be published in its present form, as it should be of interest to methodologically-orientated researchers working on similar topics. It is possible that small improvements by the authors themselves or others could find ways enhance the practical usability of the methodology presented here. In my view, a 3-D extension of the method, capturing high-rise buildings as well as single-storey ones, would provide population density information and thus introduce the human component that is missing from the present study.  

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This submission is a very interesting paper with a very applied topic and falls in the scope of the journal: International Journal of Geo-Information. Although, it has some potential for improvement. 

1. The general description of the problem (Introduction) and the description of its importance for the science and the society could be further improved. I believe that the authors should establish why the topic of this work is quite important and needs further investigation.
2. The degree of innovativeness of the methodological approach is not convincingly demonstrated. Some more details about its innovative features could further improve the quality of this paper. For example, why is this paper likely to be cited in the future?
3. Indicating the current gap in knowledge and possible limitations, will further improve this paper.
4. A greater  review of the current state of the art would further improve this work.
5. Please explain in introduction as well as in the discussion sections the contribution of this work in the relevant international literature.
6. What kind of research gap do you cover? 
7. A bit more text regarding the originality of this work and why it contains new results that significantly advance the research field.
8. I believe that adding a bit more text on why the results of the method are satisfactory (evaluation approach) will increase the quality of this work
9. Could the results be more satisfactory if you have changed something in the methodology?
10. Are the results (or the method) sensitive to this specific study area?
11. In the Discussion section I would have wished to see more information on the actual meaning of the findings and how the results add to the broader topic as well as the specific scientific field
12. The "Conclusions" section, could be further improved by describing the importance of this work, the highlight of potential further development of this methodology.

Overall I believe this is a good paper. It needs some more extra details in order to be more readable and scientifically coherent. It is an interesting paper that I would happily read, cite and suggest to other colleagues.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

The paper presents a novel method to analyze and characterize urban complexity using building footprints and fractal theory. Overall the paper is well-written, and the methods are clear. I only have a few smaller questions:

1. The only thing which was not really clear for me, is how "neighborhoods" were defined. Was this an existing analysis unit based on administrative units or was it some kind of other aggregation compared to the street block level?
2. Was there any filtering applied on the OSM street network layer or every type of street segment was considered?
3. From a conceptual point of view does that make any sense to consider anything else beyond the street network (and of course the building footprints) when defining units or calculating the index? For example topography or land use? Or these things are more important only in the interpretation phase?
4. It was mentioned that there is a tool in preparation. Will it be openly available? If yes, in what format? (code, or even as a module for GIS softwares?)
5. Is it planned for future works to compare different cities (and not only different areas within the same city)? Are the GFI values in that regard "absolute"or rather relative. For example if applying it to a city in the US with perfect grid layout, will the values be significantly lower compared to the areas in London with low values (but not regular grid-like street blocks)
6. Is there any way to perform this characterization at different scales (hierarchies)? Or it won't make more sense at higher scales than neighborhood. So for example, when comparing cities, can you assign a single value to the whole city based on the structures? 

Author Response

Please see the attachment.

Author Response File: Author Response.docx