Quantification of Annual Settlement Growth in Rural Mining Areas Using Machine Learning

Round 1

Reviewer 1 Report

The authors present a method for quantifying annual settlement growth in mining areas in Burkina Faso. 

 

The manuscript is well-written and easy to follow. The introduction provides sufficient relative studies, the methodology is well-described and the experimental results are convincing. 

 

I have a few minor suggestions:

 

1) It would be good to include two more flowcharts. One for the data acquisition and training data preparation. And one for describing the steps of the post-classification analysis. This way, it would be easier for the reader to understand the overall methodology.

 

2) Please include axis labels in Figure 4.

 

3) You could add one last paragraph in the introduction to summarize the contribution of your study.

 

4) I would be good to elaborate more on the input of SVMs (and in general of all classification algorithms that you were used). What is the dimension of the input and what does each dimension stand for. 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Line 11 and 12:  Reword to make sentence more clear.

Line 20: Take out "In" before "humid" and capitalize "Humid".

Line 49:  change "tracing" to "trace"

Line 53:  add "few" after "over the last".  Add a comma after "decades".  Thus, "over the last few decades,"

Line 63:  Reword to "But at the 30m pixel size".  Add comma after "imagery provides"

Line 64:  Remove comma after "complexes" at end of sentence

Line 94: Add comma after "mining area"

Line 102:  Add comma after "classification"

Lines 106 and 107:  Remove "from" two times

Line 108:  Take out "As" at beginning of sentence.  Add "comprised of" after "training data"

Line 129:  Add comma at end of sentence after "heights"

Lines 141 and 142:  Where pixels also used here?  This is not clear.

Line 211:  Add comma after "rural areas"

Line 217:  Sentence needs rewording to be clear.... "varied degree of variability"

Line 219:  Add comma after "Tarparko)"

Line 226:  Add comma after "consistency correction"

Figure 4:  What does the data before the red line for Bissa mean?  Why is it useful to see if there was no mine there during that time?

Figure 4:  Why are there 2 grey lines on some figures?

Line 276:  Three aspects were mentioned but I did not see/find where the three aspects were mentioned or discussed in the paragraph.

Paragraph 2.3.2 Image Classification noted five different methods that were tested.  I found no where in this entire document where the implementation or performance of these methods were discussed.  This should be core to the paper.

General:  The performance metrics were very poor for almost every case.  The greatest finding in the whole paper can be found on Lines 327 and 328 which is, the methods noted in this study can be used in "humid climates where high quality satellite imagery exists".  I am not sure if this is novel information.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

The manuscript is well written and the approach proposed by the authors is interestingThe suggestions I have to improve the text are largely minor, as detailed below.
- Line 85: what do you mean by “comparison areas”? What is their role exactly?
- Line 91: why 10-km buffers?
- Lines 91-92: the word “town” is repeated.
- Line 106: define WRS.
- Lines 106 and 107: this sentence is difficult to read, please add some commas to separate sensor identifiers and band numbers.
- Lines 109-113: What was the aim using two approaches?
- Line 145 and beyond: “SVM” should be written in upper case.
- Lines 144-147: please provide more detail on how these preliminary experiments for classifier selection were carried out.
- Line 153: was the visual assessment always performed by the same person? If not, did the authors use any measure to avoid inconsistencies?
- Line 155: how many scenes per year were available?
- Lines 180-181: this sentence is unnecessary.
- Table 1: the number of images discarded is quite high. Could you identify the main reasons for image rejection?
- Lines 272-274: are you certain that this lack of differences in settlement growth patterns was mainly due to the lack of accurately classified areas? Isn’t it possible that there are indeed no significant differences between mining and comparison areas, due to some unforeseen reason?
- It is repeatedly stated in the manuscript that results using spatial consistency checks were superior, but no hard data supporting this claim is provided. Please provide some results comparing the performance with and without the check and also show some evidence that applying the temporal check “may lead to an underestimation of the built-up areas”.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

I read each reply comment and am happy with what was recommended/added.