Modelling Spectral Unmixing of Geological Mixtures: An Experimental Study Using Rock Samples
Round 1
Reviewer 1 Report
Dear Editor,
In this paper, the Spectral unmixing of geological mixtures focused on the serpentinite and granite were conducted with the multiple model such as the linear, bilinear, polynomial models, SVMs and HSN. The fractional abundances of the endmembers were estimated by the XRD analysis. The HSN-based approach yielded reduced errors of abundance estimation, image reconstruction, and mineralogical composition for serpentinite and granite. The motivation and method was interesting. However, the organization and description of this manuscript should be minor revised before accepted.
(1) The format of this paper, especially the size of the figure and table should be critical revised. There are large blank areas in some pages.
(2) The spectra were obtained in the lab condition. Acturally, the amplitude, range, peak-value and the valley depth of the spectrum for different types of the mineral were affected by multiple factors, such as the graininess, observarion angle, roughness etc. It is suggested that if the authors could provide the discussion on the influence factors in this paper.
(3) The rock samples of serpentinite and granite were selected and analyzed in this paper. It is suggested that more typical rock samples with different mineral component can be considered in the further studies for better application.
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
The paper compares several spectral unmixing models to verify their capability to predict abundance of geological mixtures’ endmembers, concluding that HSN models give the best agreement with ground truth among the considered models.
The manuscript is clear and well written and gives useful results for planetary and geology spectroscopists. I recommend publication after the following minor revisions:
· The mixture described at lines 50-54 is the areal mixture, that described at lines 55-59 is the intimate mixture. Please add these terms, in order to facilitate comparison with terms used in literature.
· Line 74: in the works BY Altmann
· Line 178: Please summarize the Pixel Purity Index algorithm
· Line 215: Figure 6 should be moved here and named as Figure 4
· Section 4.2: Please define ARE
· Section 4.2: Which is the threshold above which RMS and ARE are considered not acceptable? Are those indicated at line 378? If so, almost all methods could be accepted.
· Conclusions: According to my comment above, I would say that nonlinear spectral mixture models provide better fits, but linear models are however acceptable.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
The manuscript "Modelling spectral unmixing of geological mixtures: An experimental study using rock samples" by Mohan Sahoo et al. reports results of applying several spectral unmixing techniques to hyperspectral data of two samples, serpentine and granite rocks. In general, I find the manuscript to be well written and interesting, and I think the reported results can be published more or less as they are presented in the current manuscript.
The experimental setup here is a laboratory setup with fixed angles. Generalizing the results into real-life situation with varying illumination levels and incidence and emergence angles would not be straightforward, and this limits the general impact of the results. However, in these controlled conditions the results are interesting and seem to favor the half siamese network.
There are a lot of results generated using five different unmixing methods. The authors do not really discuss if they were using some libraries for these methods or programmed them themselves. I think this information could be useful for the readers.
Two very small comments -- in top of page 4 you mention sensor FoV, but it would be nice to tell what is that number. In Fig. 2, it would be nice to either see a size bar or read from the caption how large areas are imaged there.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
