Urban Green Connectivity Assessment: A Comparative Study of Datasets in European Cities

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1. The novelty of the paper is not demonstrated by a deep analysis of the state-of-the-art; in other word, I am aware about many other studies on the same topic that have not been considered by the authors. In fact, the authors only cited one paper published in 2023, which can be inferred that the authors' grasp of the latest developments in the topic may be inaccurate.

 

2. Although the authors provided 10 keywords, some of them are far from the core of the manuscript. For example, there is a significant difference between Urban Green infrastructure and the Urban Green Areas. Spectral similarity is not closely related to the manuscript. The authors need to further organize the keywords.

 

3. The authors mentioned in the Abstract and Introduction that evaluating urban landscape connectivity indicators is often plagued by outdated land classification. However, what is confusing is that the authors used Sentinel-2 data from 2017 in the manuscript. Why didn't the authors use the latest data?

 

4. The authors used NDVI to evaluate the connectivity of urban green spaces. Although NDVI is a direct and useful spectral index, more and more studies have shown that NDVI inevitably comes with varying degrees of saturation. The authors need to elaborate on the reasons for choosing NDVI and analyze its sensitivity in urban connectivity. How to solve the limitation in urban green space connectivity?

 

5. The Discussion section can be appropriately divided into subheadings.

6. Page 128, “In urban… -1 to +1”. This sentence is confusing. According to the formula of NDVI, its value range is between -1 and 1 anywhere, rather than in urban landscapes.

 

7. Page 233. Please change “Normalised Difference Vegetation Index (NDVI)” to “NDVI”. In addition, the full name of NDVI is Normalized Difference Vegetation Index, not Normalised Difference Vegetation Index, please check the entire manuscript.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors, I've read your paper that I've found very interesting. I have some simple remarks for you:

- the choice of the cities was made by considering their heterogeneity at different scales but it would be suitable specify the total amount of cities that you examine and why the choice falls on these cities.

- NDVI is the most common spectral index used on RS. Considering the target of your paper maybe it could be useful to consider also other SI for manmade land cover types (for a better understanding of UGA connectivity)

- Reflection: based on your final consideration and on previous suggestion, to ameliorate UGA analysis for each data, an image classification based on SI or supervised classification could be helpful to obtain specifically land cover maps (detailed). At the same time, as you mention, time series analysis can improve a local to regional trend analysis. Maybe such kind of research could be considered for the future.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

This article analyzes urban green area (UGAs) distribution and connectivity across seven European cities using two different datasets: a classified land cover dataset and an NDVI dataset derived from Sentinel data. The authors note that they wish to address how dataset choice affects the quantification of UGAs. Overall, the study is sound. Additionally, I think it is an important contribution to urban biodiversity studies. This is especially important given the continued urbanization trajectories.

While I actually have very few comments, I would be remiss if I did not note that the findings here are not novel. Indeed, the landscape ecology literature is replete with examples of how land cover classification schemas can influence the resultant landscape analysis using various landscape metrics. Indeed, multiple authors have noted that this is merely an extension of the well-known modifiable areal unit problem (MAUP). Several notable references I think the authors should consider are Jelinski and Wu (1996) and Hay et al. (2001). Links below:

https://link.springer.com/article/10.1007/BF02447512

https://link.springer.com/article/10.1023/A:1013101931793

This is the first area in which I think the authors could improve their manuscript. The authors mention in the discussion that there is no universally prescribed or optimal dataset for quantifying urban connectivity. And this is, indeed, true. This is where an inclusion of MAUP in the discussion could greatly improve the manuscript. And this really wouldn’t require much, in my opinion, maybe just a couple of sentences and key citations (see above).

The other issue I have is that the authors state that spectral data is “generally preferable to land-cover classifications…” I am not convinced because I think this depends entirely on the land cover classification being used. In the case of this study, the authors used a specific land cover classification that was available for all seven cities (which is completely understandable). As such, I believe their claim is specific to the dataset they used, alone. Other land cover classifications could potentially outperform spectral datasets. The authors seem to recognize this when they write “it is crucial to improve the available land-cover classification…”

The authors need to note the limitations they encountered was specific to the dataset they used. For instance, researchers (with appropriate time and funding) could produce their own land cover classifications that could potentially outperform spectral data. For instance, how cool would it be if one could use exceptionally high resolution imagery gathered from a sUAS (drone) for similar analysis? (Feel free to run with that idea if you have the wherewithal.)

Those minor quibbles aside, I want to reiterate that I thought this was a good paper that provides a valuable contribution to urban biodiversity studies. It was well written and, with some minor revisions, should be ready for publication. Beyond that, I just have a handful of specific comments:

Line 23 in the Abstract: Personally, I would cut the part of the sentence that says “This study goes beyond the state of the art.” It’s a good study, it is a useful study. But it is addressing issues Jelinski and Wu noted in 1996. It addresses an issue that was first formally presented in the 1930s. Don’t oversell it.

Line 75: This is a minor issue, but the authors use the word “compromise,” but used
compromising” several lines back. Select a different word, maybe “affect.”

Line 87: The sentences separated by “comprehensively. Landscape” could actually be end and start of a paragraphs, respectively. I.E. start a new paragraph with the word “Landscape…”

Table 1 is really formatted poorly. This may just be on my end, but please fix this for final publication if it is a universal formatting issue.

Also, per table 1, why use 2011 census data? It is now 2024. A lot can change in 13 years. Are there more recent data?

Beginning on line 234 (or so), the authors provide a background on NDVI. While this is fine, I would assume most readers of Remote Sensing are quite familiar with NDVI calculations. As such, I am not sure this is necessary.

Line 238: Similarly, the authors write that in urban landscapes, NDVI values range from -1 to 1. This is everywhere, not just urban environments. (That doesn’t mean that all landscapes will have values throughout the range, merely that this is the range of NDVI values.)

Otherwise, everything looks good.

Author Response

Please see the attachment

Author Response File: Author Response.docx