Estimation of Pb and Cd Content in Soil Using Sentinel-2A Multispectral Images Based on Ensemble Learning
Round 1
Reviewer 1 Report (Previous Reviewer 1)
1. The map of the research area lacks a scale bar.
2. Due to the fact that the input set was divided into spectral, terrain, and spatial data, I suggest describing the input set separately for clearer presentation. For example, in 2.2.3, a) spectral features, b) spatial features, and finally c) topographic features. And it is also necessary to display the abbreviations of input features here to prevent readers from confusion about SHAP analysis in subsequent chapters. Alternatively, in order to present the results more clearly, the spectral, spatial, and topographic features on the y-axis in Figure 3 need to be distinguished.
3. Line 276-285: This paragraph should describe the implementation process or mechanism of stacking.
4. Figure 2 (c): The font on the y-axis is inconsistent with others
5. Line 502: There is an extra symbol.
6. The flowchart for the stacking method was missed. Additionally, the reason why stacking's performance was not as good as blending needs to be discussed.
The English quality of the paper is acceptable, and the content can be understood.
Author Response
Thank you very much for your valuable feedback. Based on your suggestions, we have further improved this paper. Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report (New Reviewer)
Dear authors,
Here you can find the review of the manuscript remotesensing-2348052. The manuscript is conforming to the journal requirements and formats, and a suitable quality data have been included. The major recommendation the authors must be taken into account is related to the figures. I hope the following comments can help authors to improve the present manuscript.
COMMENTS AND SUGGESTIONS
Title
I recommend change the title because only Pb and Cd are analysed, not more heavy metal. Please correct it in order to substitute “heavy metal” for “Pb and Cd”.
2. Materials & Methods
- In order to justify the point source pollution, is highly recommended explain if mine tailings, mine sludge, abandoned old mines, etc are located at the study area. They could be others pollution source points. It would be normal that smelters are located close to the mines where metals were mined.
3. Results
- Table 1. SBV values need a reference.
- Figure 5. It is necessary include the values for the colour scale for Pb and Cd. It is linear, logarithmic, …?. “High” and “low” do not sound scientific.
Please, increase the lettering size in all figures.
I urge the authors not to view this solely as a critique but a chance to improve this paper to convey their ideas and data in the best possible and most accurate manner.
Warm regards
Author Response
Thank you very much for your valuable feedback. Based on your suggestions, we have further improved this paper. Please see the attachment.
Author Response File: Author Response.docx
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
Review decision: major revision
Review comments:
This paper proposed the soil heavy metal prediction models with high precision inversion. This study took Pb and Cd as the main research objects and mapped the spatial distribution of them. The study also showed the advantages of the Bleding integrated learning method compared with stacking method. However, I recommend a supplement of the difference between stacking and blending in structure. Also, the figure 3 is a little confusing, especially with the lines.
In addition, the main review comment is that the discussion needs to be rewritten. The discussion section lacks some comparison with the conclusions of other articles and there are some spelling mistakes.
Several detailed recommendations were as follows:
1. In figure 1, the map of China may be not complete.
2. In 2.1 part, the contents of cadmium has not introduced.
3. In 2.2 part, please describe how deep the soil sample came from (0-XX cm?).
4. I think the number of the chapter on 282 lines may be wrong.
5. In 2.7 part, please introduce what the author does with Kriging.
6. What does the dotted line in Figure 4 represent?
7. There are MANY problems with CASE in 3.2, please check and correct them. Also, Lines 406 and 408 have the same problem.
8. Please explain the meaning of many of the abbreviations in 3.4, such as YC D, etc.
9. Please explain why the accuracy of SVM is particularly low when predicting Cd based on spectral features?
10. The legands in figure 5 are too small.
Comments for author File: 
Comments.pdf
Author Response
请参阅附件
Author Response File: Author Response.docx
Reviewer 2 Report
The title of the paper is not clear
The organization of the paper must be improved:
- Page 7, line 282 I suppose this subsection is 2.3.2
- Section 2.4 is missing
Moreover, the description of the used methods of section 2.3 could be inserted in Appendix
Page 5, lines 213-218, I suggest to organize this discussion in a Table, or to drop it (trivial)
The sampling and laboratory analysis (section 2.2.1) is a huge task. It is not clear if it should always be performed to apply the proposed method, or if it was it done for comparison purpose.
Quality of Fig. 5 must be improved.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
This manuscript investigated combining spatial features and sentinel-2A imagery to estimate soil heavy metals content based on ensemble learning. The investigated topic is interesting. However, the research and interpretation of scientific issues in the manuscript is not sufficient, and the impact of terrain on the model is not reflected in the manuscript. Results and discussion only cited three references. The explanations in discussions need to be greatly improved. Therefore, the manuscript is not suitable for publication in current stage.
The specific comments are listed below:
Line 25: The full name should be used for the first time.
Lines 39-41: Reference？
Line 65-66: Is there any previous work on estimating soil heavy metals by Sentinel 2?
Line 94: The knowledge gap is not well present.
Line 134: Please improve the quality (resolution) of Figure 1.
Lines 141-142: What did a carefully planned sampling strategy mean? It is not clear. When did those soil samples collect?
Lines 154-155: Did the Sentinel 2A data match the time of soil sampling?
Lines 159: What was the resampling method?
Lines 367-369: Why does the impact of human activities on the environment cause the spatial distribution of Pb concentration to be more dispersed than Cd?
Line 382: Initial not capitalized. (e.g. Line 387, Line 406... It is recommended to check the full manuscript)
Line 377: The figure is missing.
Lines 384-387: The sentence is too long.
Lines 381-391: It is recommended to insert chart information for readers to read. (e.g. 3.4. It is recommended to check the full manuscript)
Lines 405: “83.805 to 28.5708”?
Line 407: ”1.0579 to 0.3169”? Please check.
Lines 410-411: “should indicate that spatial and terrain features are added”?
Lines 412-416: Recommended to merge Tables 2 and 3.
Lines 421-432: e.g. “YC_D”? “SB_D”? “WY_D” ..... ? What did these mean?
Line 425: From Table 5 that there may be a negative correlation.
Lines 442-444: Cross-page diagrams are not recommended.
Lines 451-452: Why use “SVM” and “AVI”?
Lines 463-466: After Kriging interpolation, the predicted map of heavy metals is obtained, but not the measured map. And why choose Kriging interpolation as the prediction comparison and not some other method? Needs literature support.
Lines 479-493: Description confusion. There is no direct connection between “Lines 479-482” and “Lines 482-493”.
Lines 509-510: In 3.4, the wind speed and precipitation are not mentioned in the feature. In addition, with so many topographical factors discussed in the methodology, the results show that TWI is only reflected in the importance of Cd in 3.4. Can we assume that the topographic factors in the author's article have little effect on Pd and Cd?
Lines 530-561: Too many descriptive phenomena, no explanation mechanism.
Lines 591-698: Please check the format.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
