Integrating Remote Sensing and Landscape Characteristics to Estimate Soil Salinity Using Machine Learning Methods: A Case Study from Southern Xinjiang, China

Round 1

Reviewer 1 Report

Nothing to mention

Author Response

We appreciate you very much for your review. We’ve modified our paper especially in the introduction, the materials and methods, which can be found in the revised manuscript.

Reviewer 2 Report

As far as I can consider, the evaluated manuscript represents a complex study with relevant outputs. I see the novelty in the complex definition of the experiment. On the contrary, all the applied methods are commonly known. I was for that reason hesitating between 'major revisions', and 'rejected'. I leave the final decision up to the journal's editorial board whether such research is sufficient or not.

The main weaknesses (except for the novelty) of the presented manuscript are the following:

• Quality of presentation is insufficient: too many typos appear in the text, some parts of the text are highlighted by background colour etc. Such a presentation sheds a bad light to a higher quality of the presented concept.
• It is not clear on page 4 what was the total number of sampling points. Description to Figure 1 claims: "[...] A total of 151 points were measured [...]", however the line 145 claims "[...] After eliminating the invalid points, a total of 151 sampling points were retained for further analysis". As such, it seems that the original number of sampling points was higher than 151.
• I consider information pertaining to the GPS measurements as an insufficient one. Was the conducted measurement solely GPS or were more GNSS components involved? Was post-processing employed? What was the horizontal/vertical accuracy?
• Table 1 contains unbalanced descriptions for Landsat 8 on the one hand and Sentinel-1A on the other hand.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

The paper compares different methods for estimating soil salinity based on satellite images. The paper is extensive, very well explained in all its sections. However, there is one important issue that must be stressed. The paper talks about calibration and validation. Calibration is used as a training synonym, which in AI-based methods, the term training is more appropriate. 

It is not clear whether the term validation in datasets is equivalent to validation, and therefore validation data influences training sessions and statistics cannot be obtained from validation data; or whether the term validation is equivalent to testing, which does not influence training and statistics can be obtained from testing data.

Therefore, if the term validation is correct, the authors should repeat the training and remove the validation set and use it as a test set. If the term validation is equivalent to testing, please change the name to testing.

Also, this is a minor issue. The authors use a lot of linking words to start each sentence especially in the introduction (Moreover, Therefore, On the other hand, However,In contrast, etc). Although these words are used occasionally for emphasis, when used in all sentences, the reading becomes heavy. I suggest that the authors ask themselves in each sentence whether that word is necessary, or whether the sentence still maintains the meaning without those words a,d leave only de relevant words.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Dear authors,

Thank you very much for your revisions. As the editorial decision was 'major revisions' as well, I leave aside a decision whether the presented research is sufficient or not. I see both points of view, a unique study in Xinjiang, China, on the one hand, as well as the application of only commonly known methods on the other hand.
The manuscript was modified in line with all my comments from Round 1. Mostly on the lowest acceptable level, but still. Post-processing of the conducted GNSS measurements would be better; however, I do not see it as a necessary step when taking into account the equivalent scale of the study.

Kind regards,

Reviewer 2

Reviewer 3 Report

The answers and corrections have been correct on the part of the authors. I would like to congratulate them for their good work