Spatial Distribution Prediction of Soil Heavy Metals Based on Random Forest Model
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsReview of manuscript "Spatial Distribution Prediction of Soil Heavy Metals Based on Random Forest Model". The study deals with machine learning model to study the spatial distribution of soil heavy metal content in a coastal city in eastern China. The manuscript is written well, however the study needs additional analysis and results for possible acceptance of the submission.
1. Table 1. The statistics of training and testing data should be presented separately.
2. Table 4: RMSE and MAE units need to be mentioned. The RF predictions are below the acceptance par. An R2 value less than 0.6 in training stage itself shows the modeling is inefficient. The prediction performance could not be accepted as R2 values are less than 0.65.
3. The spatial distribution maps are meaningless because the prediction results obtained are not acceptable. The R2 values greater than 0.75 is necessary.
4. The model input-output combination could be changed by using feature selection methods in order to obtained improved prediction results.
5. In figure 1, training and testing sample points must be shown separately.
6. Why only one ML model is used in the study? Can experiment with other methods to get better prediction results.
2
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Comments and Suggestions for AuthorsThe paper is well-structured, with a clear argument, and minor revisions are recommended for publication. Specific comments are as follows:
1. Lines 45-73. Do not list the research results by author, but to sort out and summarize these research progress.
2. Lines 100-120. A list of environmental indicators is needed, including the names of indicators, units, data sources or calculation methods.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
Comments and Suggestions for Authors
Line 13, 18,20: “RF” instead random forest as you already indicated on line #12.
25. Only natural?
81: see the first comment and check the whole text.
86: the information about country and region is needed.
Fig. 1. Does not present the location of the study area. Please add a map of China or the world with location.
2. Study area. Information about soil types and geology is need.
102-105: again, you indicated abbreviations before (79).
Information about soil depth sampling is needed.
Information about sampling design is needed.
Information about the source of a soil organic matter map is needed.
128. What is “VIF”?
128-129. Does it mean that you use “stepwise regression method”. I think not. So, please, eliminate this text.
130. What package?
131-132. it is debatable because error metrics from Bootstrap are biased. Nevertheless, further, you write that you applied cross-validation (70/30). So, please, eliminate this text “The RF model, based on Bootstrap, performs sample-with-replacement sampling, generating a control set that is not selected, thereby obviating the need for cross-validation.”
170-172. The first sentence needs to be deleted as you already mentioned it before. The second one is need to place #4.2. section.
171. what is it SPSS?
The table #2 is very strange! How correlations can be significant with R=0.2, 0.1 and even -0.056?????
Moreover, you have just a correlation between As content and organic matter and elevation. Also some correlations between Hg with matter and elev. Other “correlations” are very low… or even absent. So, almost all text in section 4.2. not make sense.
229-323. That is information for the discussion part.
Table 4. Please indicate units for RMSE and MAE (mg/kg).
254-255. I don’t understand it. First, it needs to be written in the “methods” part.
Second, usually, spatial predictions from an RF model are performed from covariates within Python or R. Its unclear and needs to describe in detail the process of generating maps.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsWhile the revised manuscript addresses many of my comments, I am still not confident in the validity of the modeling results. In the limitations section, the authors should discuss the factors that hindered further improvement in modeling accuracy.
Enhance the literature review by including recent literature reviews. The following suggestions must be incorporated:
Bai, B., Xu, T., Nie, Q., & Li, P. (2020). Temperature-driven migration of heavy metal Pb2+ along with moisture movement in unsaturated soils. International Journal of Heat and Mass Transfer, 153, 119573. doi: https://doi.org/10.1016/j.ijheatmasstransfer.2020.119573
He, M., Dong, J., Jin, Z., Liu, C., Xiao, J., Zhang, F.,... Deng, L. (2021). Pedogenic processes in loess-paleosol sediments: Clues from Li isotopes of leachate in Luochuan loess. Geochimica et Cosmochimica Acta, 299, 151-162. doi: https://doi.org/10.1016/j.gca.2021.02.021
Han, X., Wu, H., Li, Q., Cai, W., & Hu, S. (2024). Assessment of heavy metal accumulation and potential risks in surface sediment of estuary area: A case study of Dagu river. Marine Environmental Research, 196, 106416. doi: https://doi.org/10.1016/j.marenvres.2024.106416
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsAll comments were taken into account by the authors. Now I recommend the paper for publication.
Author Response
Thank you very much for your acceptance on our manuscript “Spatial Distribution Prediction of Soil Heavy Metals Based on Random Forest Model”. We really appreciate your valuable comments and suggestions.
Round 3
Reviewer 1 Report
Comments and Suggestions for AuthorsThe revised version is found to be inorder. Could be accepted.
