High-Precision Mapping of Soil Organic Matter Based on UAV Imagery Using Machine Learning Algorithms
, Xiaohe Gu 1,*, Tianen Chen 1,*, Qian Sun 1,5, Sen Zhang 1,6 and Yuchun Pan 1
Round 1
Reviewer 1 Report
General comments:
1. Define SOM in the main body of the text (other than the abstract) upon its initial use.
2. Line 62-63 (revise): Remote sensing inversion method of soil hyperspectral determination based on a surface hyper spectrometer.
3. Proper Legend (COLOR SCALE BAR) in Fig 6.
4. The authors should consider adding the following dataset as supplementary files/links to public repositories so that other researchers can evaluate the performance of their models and compare it to the current study. This will add considerable value.
a. Location (X,Y,Z) and SOM data of 40 sites in a table.
b. Original UAV imagery of the Stuyd region containing all the bands used in the investigation.
The purpose of this research is to predict SOM in Northeast China by making use of high-resolution UAV remote sensing images with a centimeter-level (sub-meter) resolution using the Random Forest machine learning technique. Field data from forty different locations have been collected for modeling purposes. When compared to the performance of other machine learning algorithms, such as SVR, Elastic Net, Bayesian Ridge, and Linear Regression, the random forest algorithm's performance was found to be superior. A number of new soil indices have been proposed as an outcome of this study.
In the study, a number of indices and machine learning techniques are compared, and the authors suggest two new soil indices: NLIrededge2 and GDVIrededge2. The random forest approach demonstrates the best performance in the sub-meter UAV dataset, as was also observed for datasets with a similar structure but a grid size of more than one meter. The determination of SOM at a high resolution can assist reduce the expenses of manpower and testing, as well as improve the timeliness of data for farming methods that are driven by technology.
Evaluated nine soil indices that are responsive to the presence of organic matter in the soil. These nine soil indices are NLIrededge2, GDVIrededge2, RVI54, RVI, NDREI, B5, lgB1, lgB3, and lgB4. Based on the sensitive spectral response characteristics of SOM in Northeast China, the authors of this paper suggest two new soil indices: NLIrededge2 and GDVIrededge2. These findings are in line with those found in earlier studies, when nir and red have been utilized extensively throughout the building of soil indices.
The content of organic matter was often found to be higher in low-lying locations, with the explanations for the same. The SOM and the five bands were found to have a negative relationship with one another.
Specific improvements that the authors should consider regarding the methodology:
1. Reporting of RMSE in %. Show MBE in %.
2. May add RPIQ.
The conclusions are consistent with the evidence and arguments presented and they do address the main question posed. However, minor modifications/additions may be carried out as suggested above.
Additional references may be cited at appropriate places which are relevant to the theme of the paper. For example:
Heil et al. 2022. https://doi.org/10.3390/rs14143349
Zhang He et al. 2021. https://doi.org/10.1002/ldr.4043
Yan et al. 2023. https://doi.org/10.3390/rs15051433
Additional comments on the tables and figures:
Proper Legend in Fig 6. The color scale bar should show ranges/classes/bins with values instead of only min and max.
1. Define SOM in the main body of the text (other than the abstract) upon its initial use.
2. Line 62-63 (revise): Remote sensing inversion method of soil hyperspectral determination based on a surface hyper spectrometer.
Author Response
Response to Reviewer 1 Comments
Point 1: Define SOM in the main body of the text (other than the abstract) upon its initial use.
Response 1: Thanks for your comments. We have defined SOM in the main body of the text (other than the abstract) upon its initial use. Specifically, "SOM" was changed to "Soil organic matter (SOM)" (Page1, Line 34).
Point 2: Line 62-63 (revise): Remote sensing inversion method of soil hyperspectral determination based on a surface hyper spectrometer.
Response 2: We are grateful for the suggestion. This sentence "Remote sensing inversion method of soil hyperspectral determination based on a surface hyperspectrometer" has been revised to "Remote sensing inversion method based on hyperspectral data by portable field spectroradiometer"(Page2, Line 65-66).
Point 3: Proper Legend (COLOR SCALE BAR) in Fig 6.
Response 3: Thanks for your comments. The color scale bar in Fig 6 have changed to show classes with values (Page12, Line 387).
Point 4: The authors should consider adding the following dataset as supplementary files/links to public repositories so that other researchers can evaluate the performance of their models and compare it to the current study. This will add considerable value.
- Location (X, Y, Z) and SOM data of 40 sites in a table.
- Original UAV imagery of the Study region containing all the bands used in the investigation.
Response 4: Thank you for your professional and thorough advice and feedback. We have gained valuable insights and knowledge from it. We completely agree with your opinion, but regret to inform you that the data in this manuscript are not publicly available. However, they will be made available to readers as needed.
Point 5: Specific improvements that the authors should consider regarding the methodology:
- Reporting of RMSE in %. Show MBE in %.
- May add RPIQ.
Response 5: Thanks for your comments. We reported the RMSE in % and added the MBE and RPIQ in the manuscript (Page9, Line 290-291; Page11, Line 369).
Point 6: Additional references may be cited at appropriate places which are relevant to the theme of the paper. For example:
Heil et al. 2022. https://doi.org/10.3390/rs14143349
Zhang He et al. 2021. https://doi.org/10.1002/ldr.4043
Yan et al. 2023. https://doi.org/10.3390/rs15051433
Response 6: We are grateful for the suggestion. We added the additional references and cited them at appropriate places in the manuscript (Page8, Line 242; Page8, Line 258; Page8, Line 269; Page8, Line 274; Page18, Line 614-616; Page18, Line 633-634; Page20, Line 726-728).
Point 7: Additional comments on the tables and figures:
Proper Legend in Fig 6. The color scale bar should show ranges/classes/bins with values instead of only min and max.
Response 7: Thanks for your comments. The color scale bar in Fig 6 have changed to show classes with values (Page12, Line 387).
Point 8: Comments on the Quality of English Language
- Define SOM in the main body of the text (other than the abstract) upon its initial use.
- Line 62-63 (revise): Remote sensing inversion method of soil hyperspectral determination based on a surface hyper spectrometer.
Response 8: We are grateful for the suggestion. (1) We have defined SOM in the main body of the text (other than the abstract) upon its initial use. Specifically, "SOM" was changed to "Soil organic matter (SOM)" (Page1, Line 34). (2) This sentence "Remote sensing inversion method of soil hyperspectral determination based on a surface hyperspectrometer" has been revised to "Remote sensing inversion method based on hyperspectral data by portable field spectroradiometer" (Page2, Line 65-66).
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The research is worthy of publication as the manuscript documents some newly recognized aspects of the high-precision mapping of soil organic matter (SOM) spatial distribution through remote sensing. The topic is suitable for the journal and of relatively broad international interest in the field of machine learning for digital soil mapping. The methodology is stated adequately and explicitly. Presented data support sufficiently the results of the study; however, the quality of their interpretation may require some improvement in terms of precision and brevity.
General comments about findings that need revision:
The title of the paper is adequate, the keywords are appropriate, however abbreviations should be avoided, and the abstract is informative. The paper is properly organized and generally easy to read/understand.
The experimental methods are not sufficiently described. Important information is not provided, such as e.g. soil classification according to WRB (2015). The experimental design had some faults in my opinion, e.g. the content of SOM was determined by potassium dichromate-external heating method.
Tables are useful and of sufficient quality. Title of Table 1 should be change.
Figures and tables should be described in a way that allows interpretation without text in the manuscript. For example, what does it mean ‘DSM’ - Fig. 1, 'SOM' - Fig. 3, Fig. 4 - unit of SOM is needeed ? Detailed notes are included in the manuscript.
Discussion chapter is weak and chaotic. Many results are not explained and commented at all.
Conclusions are not reported enough, therefore the reader cannot evaluate the data.
The references cited are relevant and up-to date.
Although I'm not a native English speaker I suggest that paper still needs some revision, particularly with respect to style, syntax and brevity.
Detailed comments are provided directly in the text.
Comments for author File: Comments.pdf
Author Response
Response to Reviewer 2 Comments
Point 1: The experimental methods are not sufficiently described. Important information is not provided, such as e. g. soil classification according to WRB (2015). The experimental design had some faults in my opinion, e. g. the content of SOM was determined by potassium dichromate-external heating method.
Response 1: Thanks for your professional comments. We have gained valuable insights and knowledge from it. (1) The details of the experimental methods have been added to the manuscript, providing a sufficient description (Page3, Line 141-149; Page7, Line 223-233; Page8, Line 240-250; Page8, Line 251-274; Page9, Line 290-291; Page9, Line 311-314; Page10, Line 315-318; Page13, Line 423-441). (2) The soil classification according to WRB (2015) has been added to the manuscript (Page3, Line 131-132). (3). We have adopted a general method for SOM content determination, which has also been reported in several articles. For example: Zhu, CM et al. 2020. https://doi.org/10.3390/s20061795; Wu, QL et al. 2017. https://doi.org/10.1007/s11769-017-0875-9; Xin, K et al. 2018. https://doi.org/10.1071/MF17101. We are grateful for the suggestion. Moving forward, we plan to incorporate additional SOM determination methods to further enhance our SOM monitoring technology.
Point 2: Tables are useful and of sufficient quality. Title of Table 1 should be change.
Figures and tables should be described in a way that allows interpretation without text in the manuscript. For example, what does it mean ‘DSM’ - Fig. 1, 'SOM' - Fig. 3, Fig. 4 - unit of SOM is needed? Detailed notes are included in the manuscript.
Response 2: Thanks for your comments. (1) The title of Table 1 has been changed to “Descriptive statistics of soil organic matter (SOM) for the 40 ground truth soil samples” (Page4, Line 164). (2) The figures and tables have been described in a suitable manner (Page4, Line 150-152; Page4, Line 164; Page5, Line 177; Page6, Line 203-204; Page9, Line 301-302; Page10, Line 331-332; Page11, Line 369).
Point 3: Discussion chapter is weak and chaotic. Many results are not explained and commented at all. Conclusions are not reported enough, therefore the reader cannot evaluate the data. The references cited are relevant and up-to date.
Response 3: We are grateful for the suggestion. The discussion and conclusion have been elaborated with more details to fully describe the methods and results of the experiment in the manuscript (Page13, Line 423-441; Page14, Line 445-460; Page14, Line 477-488; Page14, Line 490-494; Page15, Line 509-516; Page15, Line 530-546).
Point 4: Although I'm not a native English speaker I suggest that paper still needs some revision, particularly with respect to style, syntax and brevity.
Response 4: Thanks for your comments. We invited three native English-speaking experts to thoroughly and meticulously review and refine the grammar, sentence structure, and logical coherence of the manuscript.
Point 5: Detailed comments are provided directly in the text.
Response 5: Thank you for your professional and thorough advice and feedback. We have gained valuable insights and knowledge from it. We have carefully implemented the changes you suggested in the PDF file and made detailed revisions accordingly (Page1, Line 13; Page1, Line 21; Page1, Line 30-31; Page1, Line 34; Page3, Line 131-132; Page4, Line 151-152; Page4, Line 164; Page9, Line 300-302; Page12, Line 388-389).
Author Response File: Author Response.pdf
Reviewer 3 Report
The entitled "High-precision mapping of soil organic matter based on UAV imagery using machine learning algorithms" presented a research of soil organic matter inversion using UAV multi-spectral images and machine learning approaches. The mansucript is in good shape, the data is rich, the method was logical, and the result was reliable. The manuscript needs minor revisions as only some small tips needed to be handled.
For specific comments:
1. The machine learning appraoches (state of art) based UAV remote sensing applicaitons in agriulture, ecology can be added as a seperate paragraph.
2. Improve the resolution of figure 1.
3. Line 207 to line 217 can be deleted as these indicators were commonly applied ones.
The English Language reads well and it only need to improve the section abstract and conclusion.
Author Response
Response to Reviewer 3 Comments
Point 1: The machine learning approaches (state of art) based UAV remote sensing applications in agriculture, ecology can be added as a separate paragraph.
Response 1: Thanks for your comments. We have added a separate paragraph to elaborate on the state-of-the-art machine learning approaches used in UAV remote sensing applications for agriculture and ecology (Page3, Line 100-115).
Point 2: Improve the resolution of figure 1.
Response 2: We are grateful for the suggestion. We have improved the resolution of figure 1 (Page4, Line 150).
Point 3: Line 207 to line 217 can be deleted as these indicators were commonly applied ones.
Response 3: Thanks for your comments. We completely agree with your opinion and have deleted lines 207 to 217 accordingly (Page8, Line 285).
Point 4: Comments on the Quality of English Language
The English Language reads well and it only need to improve the section abstract and conclusion.
Response 4: Thank you for your professional and thorough advice and feedback. We have gained valuable insights and knowledge from it. The abstract and conclusion have been elaborated with more details to fully describe the methods and results of the experiment in the manuscript (Page1, Line 23-26; Page13, Line 423-441; Page14, Line 445-460; Page14, Line 477-488; Page14, Line 490-494; Page15, Line 509-516; Page15, Line 530-546).
Author Response File: Author Response.pdf
