Exploring the Influencing Factors in Identifying Soil Texture Classes Using Multitemporal Landsat-8 and Sentinel-2 Data
Round 1
Reviewer 1 Report
The article assesses RS-based analysis of soil properties which is not a novel topic, nor is the comparison of the performance of models, although the authors have tried to employ ML techniques for a less studied subject: soil texture.
The article has been prepared well and is informative. Yet, there are a few critical technical issues outlined below.
Technical remarks:
It is obvious that seasonal changes can impact vegetation reflectance in response to their growth cycle, air humidity, etc. So, the most crucial question is why vegetation indices (e.g., NDVI) were used proxy for assessing soil texture in a “densely vegetated region” (Section 4.1), while there are other techniques and datasets (Radar) that can be used directly. Pls explain it to readers.
Some of your results are obvious for RS scholars: Line 443: “The ideal classification accuracies usually were obtained at fine and medium modelling”. More compelling reasons are needed, especially for soil texture analysis.
There is no indication of up-scaling methods in this case study, located in a small basin, for the larger areas.
Lines 155-158: How would it be possible for a region with an annual rainfall of 1224 mm that no ‘cloud’ noise be recorded?
Lines 155-158: How are you ensuring that soil samples taken in 2012 could be acceptable references for imagery taken from 2019 onwards, given the high land use change and agricultural activities in China?
Figure 2: How could you identify vegetation coverage/type for controlling ground-truthing imagery?
General:
The manuscript needs an English edit. I have given a few examples:
Title: ‘Influencing factors’ is correct.
Line 36: “Precise spatially referenced soil information”: Such a lengthy statement is incorrect.
Line 63: “satellite products recording crop development”
Line 103: “Recently, super learner, …., is also attractive”
Lines 159 AND 162: “These …”. What are you referring to as “these”?
Section 2.4.1: Provide full wording for SVM, RF, GBDT here.
Line 405: Similarly NOT similar.
Authors must provide a few references from 2022 as they want to justify their findings.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The paper is well structured, the material and method section sufficiently describes the experimental protocol. The references are adequate to the subject and recently. I find the results and discussions clear and concise. I do appreciate the volume of the data used and also the understanding that the choice of sensors, modeling resolutions and techniques significantly affected the outputs. Therefore, the discussions and conclusions have been accordingly.
Author Response
We appreciate you very much for the positive comments on this manuscript.
Reviewer 3 Report
I enjoyed reading this paper. It is well written. I gave a few comments to improve the manuscript.
General comments:
1. I agree that Dataset C showed better overall accuracy (OA) than Datasets A and B. However, you see the differences in OA (0.80, 0.81, and 0.83) for 10 m resolution are quite less. I suggest adding the statistical analysis or p-value to show whether these values are significantly different or not. Similarly, for different resolutions (10, 30, and 60), please add the p values. When we write “The choice of sensors, modeling resolutions, and techniques significantly affected the outputs”, then it becomes important to provide significant values. I would also suggest for different ML algorithms. I think that this analysis will help users with the options they have in the future.
2. There are a few grammatical errors. Please read it carefully and correct it. I just mentioned a few but there are many like this.
3. Section3.4: Maybe you should expand this section by providing the area covered by each soil texture class. It is essential because this is the final product of your analysis.
4. If possible, please share these high-quality maps with other users, so that they do not need to repeat this analysis. I am sure that you know about Zenodo and Figshare.
Minor Comments:
- Line1-2: The title does not seem correct to me Grammatically. Please write "influencing or dominating"
- Line 48: Please write “previous”
- Line 50: Please write “These researchers noticed that...”
- Line 79: “bands”
- Line 383: “techniques”
- Line 383: “prospects for” not prospect in
- Line 388: Please remove this work “naked”.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Good luck.
