Assessing the Accuracy and Consistency of Six Fine-Resolution Global Land Cover Products Using a Novel Stratified Random Sampling Validation Dataset

Round 1

Reviewer 1 Report

I have gone through the manuscript   Assessing the accuracy and consistency of six fine-resolution 2 global land cover products using a novel stratified random 3 sampling validation dataset. The author has done very good work but still this manuscript needs more attention of the author.

1-    Introduction section:  the authors mentioned very old literature review about Land cover applications. Please update these lines46-48 with the recent literature review

2-   Please increase the resolution of the figures and clarify the text in them.

 

3-   Please compare the Global accuracy assessment of six GLC products with previous studies

Author Response

Point 1: Introduction section: the authors mentioned a very old literature review about Land cover applications. Please update these lines 46-48 with the recent literature review.

Response 1: Great thanks for the suggestion. According to your comments, we have updated lines 46-48 with the recent literature review.

Point 2: Please increase the resolution of the figures and clarify the text in them.

Response 2: Great thanks for the suggestion. According to your comments, we have replaced the low-resolution figures with high-resolution figures.

Point 3: Please compare the Global accuracy assessment of six GLC products with previous studies.

Response 3: Great thanks for the suggestion. But we were unable to provide an accuracy comparison with previous studies. In this regard, we will describe its reasons from two perspectives. First, as stated in the introduction, current accuracy assessments and consistency analyses of fine-resolution GLC products are only available at a regional scale, and little effort has been paid by scientists to evaluate the performance of six available 10 m or 30 m GLC products, especially at a global scale. Therefore, we can’t find studies that used the same six GLC products to compare our accuracy results. However, there were some studies that assessed accuracy and consistency for some of the six GLC products. Differences in sampling design and classification systems for generating validation datasets make these studies not comparable to our results.

Author Response File: Author Response.docx

Reviewer 2 Report

This manuscript conducts a comprehensive accuracy assessment and consistency analysis of six GLC products with resolutions of 10 m and 30 m (GlobeLand30, FROM-GLC30, 18 GLC_FCS30, FROM-GLC10, European Space Agency (ESA) WorldCover and ESRI Land Cover). The major task for this study is the developing of an independent GLC validation dataset (containing 79,112 validation samples). This is followed by the actual assessment of the accuracy and consistency of the six GLC products using well-established metrics. The results shed lights on interesting aspects of these GLC products.

 The manuscript is well written and structured using sound scientific methodology.

 

Comments on the Methodology

Section 4.1.2 discusses the “Interpretation uncertainty” and states that “In total, 2469 validation samples were recorded as low confidence (‘unsure’ or ‘bit sure’) by interpreters.” However, the authors left the reader in the dark with respect to what action they took in this case or regarding these 2469 samples.

 

Comments on the Structure and Writing style

The term “rare land cover types” is mentioned 10 times throughout the manuscript. However, the manuscript never clearly and explicitly defined, explained, or provided examples to what is meant by this term. I suggest that the authors provide such information first time it is mentioned (line 161).

 

I suggest moving the text on lines 212-215 to right after “Eq. (3)” on line 205.

 

Line 347 (Table 5): kappa was not mentioned as one accuracy measure that would be used (Section 3.2), nor is it defined in the manuscript.

 

Comments regarding typographical errors:

Line 168: change “focused” to “focus”.

Line 202: remove the duplicate “)”.

Line 347 (Table 5): remove the last row in the table (kappa). See comment above.

Lines 458 and 459: the term “Num” in Table 6 is different from the term “count” used in Table C.1 although they both mean the same thing. I suggest using only one of them for consistency.

Line 470: Saudi Arabia is not in northern Africa.

Line 530: remove; it is redundant considering that the symbols/colours explained in the first column of the same table.

Line 539: change “FROM-GLC10” to “FROM-GLC30”.

Line 620: I was not clear on the meaning of “performed” here, shouldn’t it be “preferred”.

Line 651 (Figure 14): remove the enlargement (zoom-in) of Shngqiu on the lower left corner of the figure as it is never discussed in the text, nor is it relevant.

Line 653: following the previous comment, remove “In the case, the high resolution imagery came from the Google Earth Pro.”

Lines 670-671: confusing sentence; some words may be missing.

Author Response

Comments on the Methodology

Point 1: Section 4.1.2 discusses the “Interpretation uncertainty” and states that “In total, 2469 validation samples were recorded as low confidence (‘unsure’ or ‘bit sure’) by interpreters.” However, the authors left the reader in the dark with respect to what action they took in this case or regarding these 2469 samples.

Response 1: Great thanks for pointing out the issue. SRS_Val dataset containing 79112 samples was recorded as low confidence by duplicate quality-controlling. The first round was independently interpreted by 10 groups, the second round was carefully checked by five experienced experts, and the third round was comprehensively reviewed by the best experts. However, there were still 2469 low-confidence samples, which illustrated that it was difficult for anyone to determine their land-cover types by interpreting high-resolution images. But these samples were still considered to be of higher quality than GLC maps because of duplicate quality-controlling. For the processing of these low-confidence samples, collecting photos on the spot is a good way to increase the confidence of samples. Given that going to the field is time-consuming and expensive, we haven’t found a better way to determine their labels than duplicate interpretation.

Comments on the Structure and Writing style

Point 2: The term “rare land cover types” is mentioned 10 times throughout the manuscript. However, the manuscript never clearly and explicitly defined, explained, or provided examples to what is meant by this term. I suggest that the authors provide such information first time it is mentioned (line 161).

Response 2: Great thanks for the suggestion. We have added examples of “rare land-cover types” in first time it is mentioned (line 161).

Point 3: I suggest moving the text on lines 212-215 to right after “Eq. (3)” on line 205.

Response 3: Great thanks for the suggestion. According to your comments, we have moved the text on lines 212-215 to right after “Eq. (3)” on line 205.

Point 4: Line 347 (Table 5): kappa was not mentioned as one accuracy measure that would be used (Section 3.2), nor is it defined in the manuscript.

Response 4: Great thanks for pointing out the issue. According to your comments, we have deleted the text related to Kappa.

Comments regarding typographical errors:

Point 5: Line 168: change “focused” to “focus”.

Response 5: Great thank for pointing out the issue. We have corrected “focused” to “focus”.

Point 6: Line 202: remove the duplicate “)”.

Response 6: Great thank for pointing out the issue. We have removed the duplicate “)”.

Point 7: Line 347 (Table 5): remove the last row in the table (kappa). See comment above.

Response 7: Great thank for pointing out the issue. We have deleted the last row in the table (kappa) according to your comments.

Point 8: Lines 458 and 459: the term “Num” in Table 6 is different from the term “count” used in Table C.1 although they both mean the same thing. I suggest using only one of them for consistency.

Response 8: Great thank for pointing out the issue. We have unified the term “Num” in Table 6 and the term “count” used in Table C.1 as the term “Num”.

Point 9: Line 470: Saudi Arabia is not in northern Africa.

Response 9: Great thank for pointing out the issue. According to your comment, the sentence in L470 has been readjusted as: “It was found that all GLC products achieved higher accuracies in these ‘simple land-cover’ countries, such as Algeria, Libya and Egypt in northern Africa and Saudi Arabia (mainly distributed a large area of desert) ……”

Point 10: Line 530: remove; it is redundant considering that the symbols/colours explained in the first column of the same table.

Response 10: Great thank for pointing out the issue. We have removed the redundant legends of the table in Figure 10.

Point 11: Line 539: change “FROM-GLC10” to “FROM-GLC30”.

Response 11: Great thank for pointing out the issue you found. We need to explain this part. Line 539 illustrated which GLC product had higher area correlation coefficients than other GLC products. According to the calculation, we found that there was a high similarity of area allocations between FROM-GLC10 and other GLC products with an average area correlation coefficient of 0.897. And the average area correlation coefficient between FROM-GLC30 and other GLC products was 0.893. Therefore, we have made corresponding revisions based on considering your suggestion.

“It was also found that there was a high similarity of area allocations between FROM-GLC10 and other GLC products with an average area correlation coefficient of 0.897. And the average area correlation coefficient between FROM-GLC30 and other GLC products was 0.893.”

Point 12: Line 620: I was not clear on the meaning of “performed” here, shouldn’t it be “preferred”.

Response 12: Great thanks for the suggestion. According to your suggestion, the sentence in L620 has been readjusted as: “2) some land-cover types such as impervious surfaces and wetlands usually showed the fragmented spatial structures and variable and diverse spectral characteristics.”

Point 13: Line 651 (Figure 14): remove the enlargement (zoom-in) of Shngqiu on the lower left corner of the figure as it is never discussed in the text, nor is it relevant.

Response 13: Great thank for pointing out the issue. We have removed the enlargement (zoom-in) of Shngqiu in Figure 14.

Point 14: Line 653: following the previous comment, remove “In the case, the high-resolution imagery came from the Google Earth Pro.”

Response 14: Great thank for pointing out the issue. We have revised the part based on your comments.

Point 15: Lines 670-671: confusing sentence; some words may be missing.

Response 15: Great thanks for the suggestion. the confusing sentence has been readjusted as: “Overall, the details of the thematic definitions of same land-cover types are substantially different in the six GLC products, resulting in differences and uncertainties among GLC products that cannot be eliminated.”

Author Response File: Author Response.docx