Global 10 m Land Use Land Cover Datasets: A Comparison of Dynamic World, World Cover and Esri Land Cover
, David N. Barton 1, Tirthankar Chakraborty 2, Trond Simensen 1,3 and Geethen Singh 4,5
Round 1
Reviewer 1 Report
The article compares the accuracy of three of the major global land-use/land-cover (LULC) classifications to help inform future researchers on which classification product may be most suitable for their research question. In my major comments I note three concerns that may limit the contributions of this article if left unaddressed.
Major comments:
11. The abstract states: “we could not be sure that the validation data we used were independent of the data used to train the Esri deep learning model”, but (as far as I could find) this lack of independence is not robustly discussed in the methodology or the results/discussion.
22. The comparability of the classes across the three LULC classifications is not sufficiently discussed. As a starting point, table 1 should include the definitions of each of the land cover classes as defined by the metadata of the classification. Additionally, the definition for each one of these classes used by the validation data should also be included and compared to the definitions for the 3 LULC classifications. Lines 288-289 & 313-322 give are examples of the differences between the classes and the implication for result interpretation. These differences must be acknowledged from the beginning.
33. The accuracy measures discussed from lines 225-240 are hard to follow. Here is the format I suggest for helping readers quickly understand the differences in overall accuracy:
|
|
Esri |
DW |
WC |
|
Global Validation |
% |
% |
% |
|
European Validation |
% |
% |
% |
I also recommend making the accuracy numbers bigger in Figures 4-6. They are one of the most important parts of the figure but the hardest to see.
44. Lines 360-362 seems to be an important point because it also repeated in the abstract but needs more explanation. What is “design-based inference”? How does it differ from the other recommendations for users touched on below (lines 360-386)? Similarly, lines 397-399 need attention, it is not clear what this sentence means.
Minor comments:
11. Abstract seems quite long, think about what could be cut to make the main points of your findings come across more clearly.
22. The information presented in the paragraph between lines 82 and 95 would be better presented as a table.
33. Lines 122-123: Explicitly state what the four classes are that have been aggregated.
44. Line 136: Needs to be made clearer that this hexagonal grid was applied globally. Please also include a justification/citation for why that size hexagon was chosen as results will differ by hexagon scale (Goldblatt et al. 2018).
55. Line 150: Please include a discussion of how well the validation data was distributed across the different classes (this fits in with major comment 2 above).
66. Figure 3: It could be the version I have for review, but the resolution of Figure 3 was too low to interpret the figure.
77. Line 390: Inference is spelled incorrectly.
Reference:
Goldblatt, Ran, Michelle F. Stuhlmacher, Beth Tellman, Nicholas Clinton, Gordon Hanson, Matei Georgescu, Chuyuan Wang, et al. “Using Landsat and Nighttime Lights for Supervised Pixel-Based Image Classification of Urban Land Cover.” Remote Sensing of Environment 205 (February 1, 2018): 253–75. https://doi.org/10.1016/j.rse.2017.11.026.
Author Response
Please see the attachment
Author Response File: 
Author Response.docx
Reviewer 2 Report
Authors tend to present a comparison of dynamic World, World Cover and Esri Land Cover. And comments are listed as follows.
1 Authors should strengthen description of novelty of the this manuscript in the part of abstract and introduction.
2 Authors should present flowcharts of methods in Section 2.
3 What’s the purpose of comparison among dynamic World, World Cover and Esri Land Cover?
4 What are applications of the comparison in this manuscript?
Author Response
Please see the attachment
Author Response File: Author Response.docx
Reviewer 3 Report
The authors present paper Global 10m land use land cover datasets: a comparison of Dynamic World, World Cover and Esri Land Cover. The manuscript is well-written and nicely presented, as well.
However, I'm not sure that the last part of the Discussions, Potential for future research, is really necessary. It could be moved to the appendices.
Furthermore I mention that I was lucky to review a very good paper.
Author Response
Please see the attachment
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
I think this paper has been revised carefully according to the comments from reviewers. This manuscript can be accepted for publication in this version.
