A Comparative Study of Water Indices and Image Classification Algorithms for Mapping Inland Surface Water Bodies Using Landsat Imagery
Round 1
Reviewer 1 Report
Please find attached comments on your manuscript.
Comments for author File: 
Comments.pdf
Author Response
Please see the attached file.
Author Response File: Author Response.pdf
Reviewer 2 Report
Given this is a resubmission, the authors have adressed my previous concerns. The manuscript quality is significantly improved and tackles questions of not only accuracy amongst each method but also computational efficiency. With a careful proofreading, i can now recommend the manuscript for publication.
Author Response
Thanks for your comments and suggestions.
Reviewer 3 Report
The authors have addressed my previous concerns in a decent way and in my opinion might be published in remote sensing.
Author Response
Thanks for your comments and suggestions.
Reviewer 4 Report
Dear authors,
I have attached my suggestions in the file attached below.
Comments for author File: Comments.pdf
Author Response
See the attached file.
Round 2
Reviewer 1 Report
Greetings,
A detailed response to reviewers was provided and several changes have been made. I am pleased to see that the main issues have now been addressed (abstract, conclusions, some discussion) and the extra section (regression curves) has rightly been removed here.
The errors WIxx have usefully been changed to MNDWIxx, AWEIxx in the text. Though it may be difficult to change this in all figures and tables in supplementary materials, it would be very helpful for readers to change this in Table 4 at least.
KNN should be defined in the abstract.
In section 4.3, and in conclusions over the performance with TR over SR, it would be good to highlight that this research was carried out on single dates and lakes (with limited/no turbidity, vegetation issues). Index and SR/TR performance will differ when detecting flooded areas over successive images, and as water levels reduce and mixed water (vegetation/turbidity) issues increase.
Please check Figure 2, this does not display properly in the PDF file.
In future work, I recommend providing omission and commission errors directly. Relative error provides interesting information on the percentage error but it suffers from not being pixel-based. A relative error may show an underestimate of for example 10% or let’s say 300 pixels. A confusion matrix may however show that this is the result of an omission error of 400 pixels and a commission error of 100 pixels elsewhere. These balance out to provide an underestimate of 300 pixels when actually the omission error is 30% higher at 400 pixels. The overall accuracy metric (or overall error as used here) though pixel based does not provide this level of detail neither, which can be valuable to readers.
Best wishes with your work.
Author Response
Thanks for your comments. Our response is attached here.
Author Response File: Author Response.pdf
