Optimized Lithological Mapping from Multispectral and Hyperspectral Remote Sensing Images Using Fused Multi-Classifiers
Round 1
Reviewer 1 Report
61: define SVM
75: define and defend “poor spatial resolution”. Is worldview 3, with ~5m multispectral resolution in the SWIR, classified as “poor” by your (as yet undefined) spatial resolution world?
Paragraph 665: define and discuss “image quality index”. How was it calculated??
Paragraph 665: define and discuss Shannon entropy. How was it calculated?
You have demonstrated and proved your hypothesis by testing classifiers against a detailed lithologic map, and found that the ensemble method gives better results than using any single sensor, or any single classification method. My great concern is that this leaves entirely open how this method can be used to improve poor lithologic maps, or create maps in unmapped areas. Without this extended discussion, the paper is incomplete, and is not of great use, other than as an academic study. If you can develop this part of the discussion fully, then I would consider re-reviewing the paper.
Author Response
We thank for the reviewer’s constructive comments and suggestions.
All the comments have been addressed and highlighted with the green colour in the revised version of the manuscript.
"Please see the attachment."
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Please refer to the attached file for detailed comments.
Comments for author File: Comments.pdf
Author Response
Authors would like to thank the reviewer for reviewing our manuscript and proving us crucial comments and valuable and constructive suggestions to improve our manuscript.
All the comments have been addressed and highlighted with the red colour in the revised version of the manuscript.
"Please see the attachment."
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
thank you for your revisions to the original manuscript
Author Response
Authors would like to thank you for your valuable time to review our manuscript for the publication in 'Remote Sensing' and providing critical comments and suggestions to improve our manuscript and highlighting the practical problem of geological remote sensing.
Reviewer 2 Report
The manuscript has been significantly improved. It is clear that the authors have made significant modifications and well addressed most of my concerns previously about the clarity of the methodology, the interpretation of the results, as well as geological implications. I also agree with the rationale behind pixel swapping that it is useful to try to reduce the effect of heterogeneous composition within a certain mapped lithology area by making the pixel swapping. I just have a few minor comments left.
First, I mentioned last time, that cross-validation of the results would help readers evaluate the performance of the newly proposed method, which should not be very complicated. I wonder if the authors did attempt to perform cross-validation of their results. They could have selected different training samples and test the accuracy dependence on different subsets of training samples. Given the accuracy numbers reported right now with a particular test, it is unclear if this is applicable in a broader sense.
Second, about the band depth formula in Section 2.1.2. The authors replied that these are parameters derived by themselves. But are these mentioned/used in the manuscript? If these formulations are not particular to this work, it is better to introduce with more generalized/common formulations of band depths (Since this section is supposed to be a literature review). e.g. the band depth formula BD = 1-R/Rc is quite commonly used but not the rest (BDR, NBDI, BNA). If the authors insist that these parameters are useful to this particular study, please add a statement as you say “The formulations for band depth and band ratio differ in different literatures. Therefore, we have used own formation in order to establish uniformly in the review. “.
Figure 4: It is better to add a note in the figure caption that these are not representative of the lithology in general. (All the absorption features are due to minor components of clay and other hydrated minerals from weathering rather than the lithology themselves (e.g. Quartzite with mostly quartz should not have absorption features in this wavelength range, but the spectra do show various absorptions.) Therefore to avoid misunderstanding of the figure that people take for type spectra of these lithologies, I suggest the addition of a warning note.
Also, the legends are still too small to read for 4a).
L469: What is the connection between the first law of geography and the pixel swapping? I agree with the rest of the statement about surface lithology being complicated by other surface covers, but it’s unclear to me how the first law of geography plays a role here. Is this statement written to justify the 3by3 window size? Please elaborate.
A few minor language edits:
L80: “particularly this problem become more difficult” : In particular, this problem becomes more difficult….
L626: “is independence to” : is independent from
Please check the language thoroughly.
Author Response
Authors would like to thank the reviewer for his/her effort to review and proving fruitful suggestions and comments to improve the quality of our manuscript for the publication in the “Remote Sensing”.
Author Response File: Author Response.pdf
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
