Generation and Optimization of Spectral Cluster Maps to Enable Data Fusion of CaSSIS and CRISM Datasets
and Benedikt Elser 1Round 1
Reviewer 1 Report
Please see the attachment.
Comments for author File: 
Comments.pdf
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 2 Report
Dear Author
I found his article very interesting.
Best regards
Author Response
We thank the reviewer for your review.
Reviewer 3 Report
Summary
This Article uses recently developed dimensionality reduction algorithms combined with different cluster algorithms to explore their combinations as applied to CRISM hyperspectral data for Mars, as well as multispectral data from CaSSIS from the same regions.
Overall, the combination of techniques and their evaluation is powerful and may have utility for future studies of Mars and other bodies.
But the paper is not particularly well written and could use improvements as pointed out in the line-specific comments below. After revision it is likely suitable for publication in Remote Sensing.
General Comments
- The codes for all the techniques used in the paper might be available via the references included, but it would be much more useful and help with reproducibility if the authors indicated where they got the code or pseudocode from, and if there is any novel code produced for this project it would help to include it as part of the data availability.
Line-specific Comments
Line 3: “Not merely is correlation of colour diversity maps with local morphological properties desirable but also mineralogical interpretation of the observations is of great interest, despite the low spectral resolution, because of the relatively high spatial resolution” – sentence is a mess; should be re-worded.
Line 6: “In this paper we combine the broad-band imaging…” the broad-band imaging of what?
Line 12: Should not be a space inside the quotation marks: ‘the so-called " summary products"’
Line 25: “our planetary system” what is our planetary system? The Earth? The Earth and Moon? The solar system?
Line 26: “Incomplete” in what sense? Spatial coverage? Spectral coverage?
Line 31: Spell out all acronyms at first use.
Line 34: See above. Follow academic conventions, do not make readers dig through an appendix to understand acronyms.
Line 35: What do the spectral capabilities have to do with surface coverage? Especially for HiRISE.
Lines 40-45: Many of the paragraphs in this paper are 2 or 3 sentences long.
Line 84: Spell out the reference names instead of putting them in brackets after a comma.
Line 105: This doesn’t really make sense. JHU/APL ran the CRISM mission; saying you’re using the datasets provided by them is redundant or implies there are multiple versions of the datasets and that the JHU ones are different.
Line 113: “sample an area of ≈ 9” this is not the correct symbol to use.
Line 124: “absolute units” what does this mean?
Line 131: This is not what “c.f.” means. Fix throughout.
Line 214: “between i-th data points and its first nearest neighbor.” Should either be “data point” or “and their nearest neighbor”
Line 266: “same color in the pictures” – be consistent throughout the paper with US vs. non-US spelling.
Figure 2: Do all the colors in this figure represent pixels assigned to clusters, or are there any pixels which are unassigned? It would be helpful to state this.
Table 1: It’s very difficult to see the differences by looking at a sea of numbers. I recommend either: (1) add commas to break up the numbers, (2) use scientific notation, or (3) if permitted by the journal, use color coding in the table. It may also help to sort by the average of the 3 regions from highest to lowest instead of alphabetical order.
Figure 5: The text in the legend and on the axes is too small: increase the font size.
Lines 352-368: It’s not clear to me what you’re doing here. I think you’re taking the 2D images from the CRISM spectral parameters and running them through the clustering algorithm? If so, is this being done with all the parameter maps or a subset of them?
Lines 357-360: This part in particular is not clear: “In this case, correlating the products within the selected spectral range with the help of a random forest classifier implies predicting the cluster maps.”
Correlating which products? And within which spectral range? Predicting what cluster maps?
“The data were split into training and test sets using a ratio of 0.25.” Which data exactly?
Line 362: “Our analysis shows that 4 features of the subsequent list are dominant for all three datasets: RBR, R770, BD860_2, BDI1000VIS and R1080.” Are you sure you don’t mean 5 here?
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
All comments from the review have been adequately addressed.
