Spatial Prediction and Mapping of Soil Water Content by TPE-GBDT Model in Chinese Coastal Delta Farmland with Sentinel-2 Remote Sensing Data

Round 1

Reviewer 1 Report

Dear authors,

The precise introduction needs some additional information as highlighted in the attached file. Dealing mainly with auxiliary data from remote sensing - Sentinel-2. It is appropriate to highlight the main advantages of this use compared to other satellites

The objective can be improved with additional information about remote sensing data and the study period.

In the methodology, some aspects need to be organized and it is important to expand other levels of important information for the greater impact of the work and replication of the study in other regions of the world.

Note: All suggestions are highlighted in the attached manuscript, in yellow tone.

Finally, the main criticism is that this study did not address the prediction of soil water content over time, that is, in several images in different years. In this case, remaining as the main suggestion for future studies.

Comments for author File: Comments.pdf

Author Response

We appreciate you very much for your valuable and constructive comments and suggestions on our manuscript. We have revised our manuscript according to the your comments point by point. We added more relevant references and improved the description of sections including introduction, materials and methods, discussion, reference, and all modifications were marked in red words. We also modified all the grammars in full manuscript by language editing services of MDPI (see supplement figure). Thank you for your kind attention on our manuscript.
For detailed revisions, please see the attachment “Response to Reviewer#1.doc”.

Author Response File: Author Response.pdf

Reviewer 2 Report

OBSERVATIONS TO MANUSCRIPT 2344898: The machine learning model optimized by TPE enables the 2 spatial prediction and mapping of soil water content in the 3 coastal delta farmland of China with Sentinel-2 remote sensing 4 data

The document is interesting, this study proposes a hyperparameters optimization machine learning method with great potential to predict the spatial heterogeneity of soil water content, which can effectively support regional farmland soil and water conservation and high-quality agricultural development. However, the document can be improved if some modifications are made.

Introduction

Line 49. Delete the year of the citation Filgueiras et al. (2020), it should be written: Filgueiras et al. [12]

Line 52. Delete [12]

Line 63. Delete the year of the citation Fuentes et al. (2022) [14], it should be written: Fuentes et al. [14]

Line 69. Delete the year of citation Chaudhary et al. (2022), it should be written: Chaudhary et al. [15]

Line 73. Delete [15]

Line 76. Delete the year of the citation Babaeian et al. (2021), it should be written: Babaeian et al. [16]

Line 82. Delete [16]

Materials and Methods

Line 109. Delete the point after the (Figure 1)

Line 112. Place a space after the number of hours 2289.8h, it should be written 2289.8 h

Line 113. It is not correct to write the type of soil if only its mode of formation is being indicated. The writing appears "...the type of soil is alluvial soil...", it should be written: the soils are alluvial

Line 129. Place a space after the number of hours 24h, it should be written 24 h

Line 353. Remove the parenthesis from the letter in the indication of the figure. For example, (Figure 6(a)), it should be written: (Figure 6a)

Line 370. Remove the parenthesis from the letter in the indication of the figure. For example, (Figure 7(a)), it should be written: (Figure 7a)

Line 375. Remove the parenthesis from the letter in the indication of the figure. For example, (Figure 7(b)), it should be written: (Figure 7b)

Discussion

Line 414. Delete the year of citation Liu et al. (2021), it should be written: Liu [54]

Line 417. Delete [54]

Line 417. Delete the year of citation Zhang et al. (2023), it should be written: Zhang et al. [41]

Line 419. Delete [41]

Line 440. Delete the year of citation Zhu et al. (2023), it should be written: Zhu et al. [60]

Line 441. Delete [60]

Line 452.  Delete the year and et al. of citation Liu and Shao et al. (2015), it should be written: Liu and Shao [65]

Author Response

We appreciate you very much for your valuable and constructive comments and suggestions on our manuscript. We have revised our manuscript according to the your comments point by point. We added more relevant references and improved the description of sections including introduction, materials and methods, discussion, reference, and all modifications were marked in red words. We also modified all the grammars in full manuscript by language editing services of MDPI (see supplement figure). Thank you for your kind attention on our manuscript.
For detailed revisions, please see the attachment “Response to Reviewer#2.doc”.

Author Response File: Author Response.pdf

Reviewer 3 Report

This paper investigated the capabilities of multiple machine learning (ML) models to map soil moisture using remote sensing-based factors.  The subject is relevant for Agriculture. However, a major concern is the implications of the developed ML model and findings: (i) Atmospheric/weather conditions, such as precipitation, evapotranspiration, radiation, air temperature, humidity, and wind have important effects on soil moisture variability. In the study, the training and testing of ML models were based upon the data from a single day without considering variations in atmospheric conditions. It is not clear how the ML models can be useful for other days with different weather conditions?; (ii) The findings are also applicable only to the single soil (or soil profile) depth for your soil moisture field sampling.  The study did not provide any information about the soil depth for the measured/estimated soil moisture (surface soil moisture or root zone soil moisture?). The readers will have no idea which soil depths the developed models can be useful for and how the model performance will be changed for a different soil depth. Overall, the implications of the present models and findings are very limited. These limitations should be seriously considered (or at least discussed) in this study.

 Other comments:

Line 1-3:   You should use a succinct and descriptive title (rather than a sentence)

 Line 120-131:  The soil sampling depth is not clear.

 Line 131:  Equation (1) seems not correct. It should be       Soilw= (m1-m2)/(m1-m3) *100%

 Line 145-149:  You should clearly point out that the 8 features were calculated for each Sentinel-2 band so that you got the 88 soil texture feature bands, which were then used for your PCA.

 Line 155-156:  “ the band after DIS transformation” --- > “those experiencing DIS transformation. ”

 Line 157/158:  In Figure 2b, except for MEAN and DIS, it is hard to tell where other soil texture feature bands are located. So the figure 2b must be modified.  Also from Figure 2b, why does PC1 also carry negative loadings (weights). PC1 represents intensity and should carry only positive loadings.  Please explain.

  Line 161:  In section 2.4, you did not clearly state the input variables (10 indices + 5 PCs) and the reference data /response (soil moisture measurements) for your ML models.

 Tables 2/3/4 have never been referred to in the text. At least, their first mention should be provided in the text. Also in those tables, it is not clear what value was used for each hypermeter in the present study. If the default value or a range of values were used, it should be clearly stated in text.

 Line 223-224:  It seems that the equation (8) is not correctly explained.   The  “c’” is a constant, but what quantity does it represent and how is the constant “c” determined?  Why “the sum of the difference between the constant and the true value of the data set is the smallest”? Which one (yi in the equation ?) represents the “truth” value of the data set in your case?

 Table 4:   What does the “friedman_mse” represent?

 Line 248:   Provide the reference citation for the “original TPE paper”.

 Line 266-267:    The subscript index for the predicted value should be ‘i"  in Equations (5) & (6). Also, explain what the “m” denote.

Line 314: Have -> has or had  

Line 320: In table 5, it is not clear that the evaluation results are for the Training or Testing.

 Line 315/338:    “scatter trend”  ->  scatterplot

 Line 338:  “true value” à measurements (to be consistent with the x-axis label).

 Figure 5:    Add the unit for RMSE

 Figure 6:   What is the soil (or soil profile) depth for the displayed soil moisture?

 Line 417:  Delete ‘study’

The standard English of the manuscript is not good.  Many sentences need to be reviewed for grammar correction.  In particular, some sentences are very long, and the clauses are not correctly separated/connected using appropriate conjunctions, e.g. for  Lines 110-112, 112-114, 271-272,  313-314.

Author Response

We appreciate you very much for your valuable and constructive comments and suggestions on our manuscript. We have revised our manuscript according to the your comments point by point. We added more relevant references and improved the description of sections including introduction, materials and methods, discussion, reference, and all modifications were marked in red words. We also modified all the grammars in full manuscript by language editing services of MDPI (see supplement figure). Thank you for your kind attention on our manuscript.
For detailed revisions, please see the attachment “Response to Reviewer#3.doc”.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

I am okay with the revision.