Soil Classification Mapping Using a Combination of Semi-Supervised Classification and Stacking Learning (SSC-SL)
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis study used machine learning algorithms for digital soil mapping, and the research was very comprehensive. However, there are still some issues, as follows:
1. The author chose a high score No.2 image as part of the environmental variable. Why did they choose this time period, whether it was the bare soil period or with indicators? The author needs to explain clearly.
2. For validation, cross validation may be a better method as it can consider all samples.
3. The abstract and conclusion need to be more concise, summarizing scientific laws.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsPlease see attached
Comments for author File: Comments.pdf
See attached
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsDear authors, after carefully reading your ms, I found this ms was well-written. I have provided some general and specific comments for your consideration.
General comments-
Discussion is insufficient. Please consider adding a discussion on the uncertainty analysis of this study.
Specific comments below- which I hope will be helpful to improve the quality of this ms.
L14: please provide the full name for “GF-2”
L18: what is the purpose of selecting unlabeled samples?
L28: put a space between “and” and “32.9%”, there are also same issues in the remaining texts, please check also
L30: Please give the full names before using abbreviations of SPM, LU, MRVBF, Ele
L60-70: add references
L76-79: what is the difference between soil type classification and soil type prediction?
L82: please provide the full name for FCM, the full name should be given the first time the abbreviation appears, please also check the remaining texts
L86: it should be: The aims of this study are..
L92: add references for section 2.1 Study area
Section 2.3.1: What method did you use for land use classification? What features were considered when doing land use classification? How many land use types were considered in this study area? Please add more description.
L206-232: Please add references for these models
Figure 1: please consider adding subfigure labels (a) (b) (c) for each subfigure. The caption cannot fully cover the figures, as you also include a DEM map, please revise the caption. Please also cite Figure 1 in section 2.1, and move Figure 1 under section 2.1.
Table 1: The column names are not exactly accurate. Please consider something like “full name and definition” and “Abbreviation” for columns 2 and 3. Please add explanations for R, NIR, G, B, …
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Comments and Suggestions for AuthorsTitle of the paper suggests improving the accuracy of soil mapping. The authors do not refer in this paper to any other data and methods proposed for soil mapping. It seems to me that the title should be changed to better align with the content of the article.
I suggest expanding the description of the geographical conditions prevailing in the analyzed area (lines 92-105). A more detailed description of the conditions of soil formation and functioning, along with maps illustrating their spatial distribution, will help the reader determine the spatial distribution of individual environmental factors.
There is no reference to Figure 1 in the text.
The authors frequently use the term "environment covariates" in various places (lines 121-126)… however, I believe these are more appropriately referred to as "environment variables." The authors should pay more attention to using commonly used terms in the literature.
In subsection 2.3.1 – lines 130-144, more detailed information describing remote sensing data should be provided. Providing only spatial resolution without information on the acquisition dates and spectral resolution, as well as the number of channels, is, in my opinion, insufficient.
The paper does not address in any way the description of methods for bringing individual data of different resolutions to the same resolution. I consider this a significant methodological error that prevents drawing correct conclusions.
In the discussion, the authors only refer to one dimension of analysis, which is model accuracy. Complete omission of some aspects, such as the impact of surface data such as remote sensing (RS) on soil, which covers material to a depth of about 200 cm, is noted. I also miss a description of why and where the model's precision is lowest and where it is highest – spatially. Such information would allow for a fuller understanding of the problem of automatic soil type recognition using RS data.
The authors also did not refer to the results achieved by other researchers in attempts to automate the soil recognition process and the results they achieved.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
Comments and Suggestions for AuthorsThanks for your great revision, but the full name of Ele is still missing in the Abstract.
Reviewer 4 Report
Comments and Suggestions for AuthorsI have gladly received the new revised version of the proposed article. After addressing the suggested changes, it seems to me that the article is suitable for publication in the journal Remote Sensing.
