The Spatial Prediction of Soil Texture Fractions in Arid Regions of Iran
Round 1
Reviewer 1 Report
My overall opinion is that the authors performed a comprehensive research in an important and relevant topic. It has an appropriate research plan with deep analysis, and it would take interest by readers. However, some parts need clarification or rephrasing. The manuscript needs more precise wording as well as consistency in acronyms. It calls for a thorough read-through because of lots of misprint. I suggest the manuscript for publication with minor revision. Before publication I recommend to improve the signed parts.
I have suggestions and remarks throughout the text, please see my detailed list below. Section 2.4. Auxiliary data needs completion the most. Section 2.5. Spatial prediction should be more consistent. Letter a, b, and c are not marked in the captions of Figure 3, and Figure 4. Figure 5. a, b, and c has not referred in the text. I hope that the resolution of the figures is low only because of the conversion to pdf format.
DETAILED REMARKS AND SUGGESTIONS
Line 20. (and other occurrences) percentage of sand, clay and silt
sand, clay, and silt content (in percentage)
Line 22, 24, 135. up to 2 m depth
In Global Soil Map specifications 100-200 cm depth is also required. In line 22, soil profiles were mentioned ‘up to 2 m depth’, but afterwards 60-100 cm is the deepest depth in the manuscript. I suppose soils of the investigated area are shallower (?), however, it should be mentioned somewhere (e.g. in line 136.) why 100-200 cm layer was omitted.
Line 27. make a correlation
I think we do not make a correlation, rather e.g. reveal it.
Line 43. Soil texture … affects the amount of water and nutrients
affects the amount of accessible water and nutrients / affects the accessibility of water and nutrients
Line 45. Soil texture is also important in the comprehensive soil classification system
Singular wording here is not completely precise, because there are more (at least two) internationally accepted soil classification systems. However, I see you refer to USDA Soil Taxonomy.
Line 49. transition functions
pedotransfer functions?
Line 59. Much research has been conducted…
Many researches have been conducted…
Here the noun is countable, because it refers to more than one separate research projects.
Line 69. scorpan model
It would be worth to mention the initial components (factors) of the acronym.
Line 71. expert knowledge
There is no doubt that expert knowledge is very important in applying the SCORPAN model, but listing it here is a bit strange, as it seems to be one of the factors.
Line 75. tree classification, tree decision
Line 82. (and many other occurrences) tree regression
Inversion of word order is more generally used: classification (or decision) tree, regression tree
Line 75. … neuro-fuzzy have been used.
If it is a general statement, the tense is not adequate.
Line 82. soil salinity map
It would be more adequate referring this method mapping of soil texture instead of salinity.
Line 86. neuro fuzzy
With hyphen at every other occurrences.
Line 93, 438. soil transfer functions
pedotransfer functions?
Line 95. The objectives of the present study are…
A new paragraph should begin with this.
Line 95-96. to predict soil texture
Incomplete phrase in this form.
Predict spatial distribution of soil texture (classes or fractions)?
Line 96. from the various particle size fractions using … models
from the various particle size fractions data
Line 98. satellite image
satellite images
Line 99. predicting clay, sand and silt
predicting spatial distribution of clay, sand, and silt content?
Line 104. The altitudes of this city have been extended in two directions from north-west to south-east
As far as I know, altitude refers to the vertical height, however, the above mentioned extensions refer to horizontal location.
Line 128. (and many other occurrences, e.g. line 154, 227) including; organic carbon (OC), pH, …
Semicolon (and any other punctuation) is not reasonable before enumeration.
Line 140. two quantitative and qualitative data were used as auxiliary data.
This sounds as if only four thematic layers were used, but there were many. Four were rather e.g. groups of auxiliary data.
Line 143. Table 2
Table 1 has not yet been referred in the manuscript up to this point. Numbering should be inverted.
Line 143. SAGA
The developers’ request: ‘Please provide the following reference in your work if you are using SAGA:
Conrad, O., Bechtel, B., Bock, M., Dietrich, H., Fischer, E., Gerlitz, L., Wehberg, J., Wichmann, V., and Böhner, J. (2015): System for Automated Geoscientific Analyses (SAGA) v. 2.1.4, Geosci. Model Dev., 8, 1991-2007, doi:10.5194/gmd-8-1991-2015.’
Line 144-146.
It would be more elegant writing the parameters consistently (I mean capital letters).
Line 149. Landsat 8 Bands 2-7
Resolution is not indicated in contrast with the other bands.
Line 158. remote sensing index were applied
indices
Line 164. air-dried soil samples
How many samples? From what depths? Do they come from the same 115 locations as samples in line 120?
Line 175. Then, the absorption values became because filtering and processing are better on it.
I think wording should be more scientific here.
Line 189. influence of the spectra on the percentage of clay, sand and silt…
On the contrary: clay, sand, and silt have influence on the spectra.
Line 189-191. Then, the influence of the spectra on the percentage of clay, sand and silt was extracted and entered into neural network, tree regression and neuro-fuzzy models as auxiliary data.
Do I understand well that this paragraph still refers to processing of the spectral data?
It seems to me, that a step is left out here. How spatially exhausted auxiliary data were generated from point data? Regression tree, ANN, and ANFIS? Later, in Results (Line 285.) Kriging is mentioned, but that is not described here at all.
Line 193. In this study, 30, 30 and 29 layers of auxiliary information were used to estimate the clay, sand and 193 silt, respectively.
If this paragraph (at least the beginning of it) refers to processing of the spectral auxiliary data, this (and the previous two) sentence(s) seem to refer the spatial prediction. However, in the following paragraph the description of the auxiliary variables is continued.
Please, consider rewriting and completing this paragraph.
Line 195. qualitative data was generate across the study area
generated
Line 196, aerial photographs at 1/40000
You mean at 1:40,000 scale?
Line 207. Three different spatial prediction approaches were employed.
Adding some opening phrases would be more elegant.
Line 208. Decision trees, such as regression and classification trees
‘Classification And Regression Trees (CART)’ is the commonly used expression.
Line 212. An Artificial Neural Network…
A new paragraph should begin with this.
What is more important: in the previous sentences there is a general description about CART. Then the manuscript suddenly turns to ANN applied in this research without general description. This section should be more consistent.
Line 222. JMP
Citation of the software is missing.
Line 232, 233, 235. close to 0, close to zero
inconsistent orthography
Line 252, 253. silt content was on average smallest of the three …, with the subsoil smallest
the smallest
Line 254. According to the US Soil Texture Triangle the average soil texture in the area …
It would be worth to present the distribution of the points on the texture triangle in a new figure.
Line 269. visible-infrared
visible- near infrared?
Line 279, 282. Unschambler
Unscrambler
Line 283, 284. For clay percent, 879 and 890, for sand percentage, 852 and 879, and 283 for the silt percentage of 1456 and 1998
Unit? I suppose nm. For percentage, see remarks for line 20.
Line 285. (See remarks at Line 189-191.) using the Kriging method
There is not any description in the methods. What kind of kriging was applied?
Line 294.
It is not completely clear for me, how the order of importance was set?
It is also not clear, what does CLAY E, CLAY F etc. mean according to Table 2.
Line 301, 302. remote sensed data
remotely sensed data
Line 321, 465, 473. close-up measurement
Do you mean proximal soil sensing?
Line 361. amount of RMSE
value of RMSE
Line 382. ANFIS
The acronym appears here first. The initial components should be decoded here, but better earlier (in the introduction or in the methods section).
Line 454. Finally, Soil texture prediction map was obtained…
Finally, soil texture class map was obtained…
This step was not mentioned in the methods section. E.g. how do you handle that the sum of clay, silt, and sand content would be 100 %? Is there any uncertainty measurement of the texture class maps?
Line 469. ANFIS, can carry out soil mapping process … with any natural complications
‘Complications’ has two alternative meanings, therefore I suggest using here ‘complexity’ instead.
Line 476. in countries such as Iran, that continuous spatial and three-dimensional data But…
Something is missing here.
Line 477. helper variables
You mean auxiliary variables?
Comments for author File: 
Comments.pdf
Author Response
The authors would like to thank the Editor and the reviewers for their precious time and valuable comments. We have carefully addressed all the comments.
Author Response File: 
Author Response.doc
Reviewer 2 Report
The article demontstrates the performance of different modelling methods for the prediction of clay, silt and sand content in arid regions of Iran. It covers regression trees, artificial neural network and neuro-fuzzy technique. The predictions are made for 5 standard depths required by Global Soil Map project.
First of all, I will strongly recommend English editing and revison of manuscript text.
References in the text are incorrect. They should be numbered in order of appearance. It means the first reference should be [1], not [3] as in line 45 of manuscript. Line 50 - reference included in round brackets but they should be [].
Another recommendation - tables and figures should be after the text where they are mentioned first time. Not before. Because it is confusing when table appears without any explanation (like table 1).
In text you mention figure 3a and etc. But actually you have 3 figures with the same caption Figure 3. In case of complex figures, they should have one common caption, and all figures in this complex figure should be named like a, b, c...It is actually illustrated in Soil Syst. Microsoft Word template file available at https://www.mdpi.com/journal/soilsystems/instructions.So, I suggest you read it very carefully and check formatting and caption of all figures and tables in your manuscript.
My next comments will be devoted to the content of the manuscript.
Firstly, the name of the article includes mention of vertical distribution, in abstract there is a mention of profiles modeled with continuous depth fucntion (line 21-22), small subsection 2.3. line 133 about spline depth equaitions, line 447 and line 474 in conclusions also mentions spline-depth equaition. However, this part is not presented in any way in results. So, you need to add this part or change the title of the article and remove any mention of this function.
Line 142 US digital elevation model is a bit confusing. If you ment SRTM, please write it.
Table 2- what is the meaning of the column Source? Because it is not clear for me when I see reference to (Boettinger et al., 2008) next to NDVI and PVI. And the reference when made should be number in [].
Earth's face data - what is it?
Line 172-175 the explanation is not clear. First you have 50 spectra that you average them into 1??? How that goes with 3 replications per sample?Do in your manuscript soil sample and soil specimen have the same meaning?
Lines 179-180 and 183-184 - duplicate information and this information is included in subsection 3.2. (lines 266-278). Strictly speaking, this should be in section Materials and Methods. So I suggest, you remove duplicates and leave this part in subsection 2.4.
line 189-190 it is not clear what you mean by the influence andhow the incluence can be used in modelling. From further reading I can suggest that you mean best weighted coeffitients (line 282). It my guess is correct, please provide clear description in mentioned lines in subsection 2.4.
line 285 - they were mapped by they you mean best weighted coeffitients? Because, it is not clear from the text.
As this is derived product, I suggest you add information on the prediction accuracy of the plsr model you considered best, maps produced by kriging together with error maps.
As far as I understood in lines 283-284 numbers 879, 890, 852, 879 etc. are wavelenghts with highest regression coeffitients? If it is so, add nm after numbers to make it understandable.
lines 319-320 - correlation is unitless and can vary from -1 to 1, so for me it is not clear what you mean in this lines.
Figure 4 - how the importance was determined and what shown numbers refer to? I understand that the higher the number is the more important the variable is, but is 3.2 means very important of just more important that 1.9
line 446 - nonagricultural profiles - first mention of them in text. So you need to provide more information on studied soils. You mentioned collecting 115 samples. Provide information on soil types and where they located (agricultural or nonagricultural land). And why did you include all profiles in modelling? Models can be different depending on land use.
line 404- what do you mean by crust? It is usually considered as compacted layer on soil surface.
Figure 6 - is of bad quality
What is the spatial resolution of all derived maps?
Lines 128-130- other properties were determined but they are not used in models or anywhere else. I do not think that this information should be included in manuscript because it creates some expectaition.
Author Response
The authors would like to thank the Editor and the reviewers for their precious time and valuable comments. We have carefully addressed all the comments.
Author Response File: Author Response.doc
Reviewer 3 Report
Dear Authors,
The idea for the study is fine and the analyses seem consistent. However, I recommend a thorough English review and resubmission because I feel that the paper can be much improved if the authors can adequately present their findings to the readers. I suggest to focus and strengthen the discussion on the methods compared as I think that the neuro-fuzzy results are the most interesting results of the study. Can you make the presentation of these results more graphical? Can the neuro-fuzzy results be detailed and properly interpreted to the reader? Please also clarify how the soil texture map was produced from the derived clay, sand and silt maps.
Author Response
The authors would like to thank the Editor and the reviewers for their precious time and valuable comments. We have carefully addressed all the comments.
Author Response File: Author Response.doc
Round 2
Reviewer 2 Report
The manuscrip has been imporved, but some corrections are still necessary.
English check should be done.
The reference in Introduction are still in the wrong order. I suggest you properly check it. References should be numbered in the text in order of appearance. Please, refer to https://www.mdpi.com/journal/soilsystems/instructions.
Table 2 (line 112) is mentioned before table 1 (line 189).
Tables should be placed after their first mention in the text.
The caption for Figure 3, 4 are incorrect. It should look like
Figure 3 Spectra of 10 selected soil sample: a) reflectance, b)......
And Figure 4 shows the importance of the variables, so it should be named properly.
Line 34 - MrVBF should be explained
Line 46 - such as ; bulk density (; should be removed)
Lines 58, 59 - have been used, have been used
Lines 169-170 modelling was done on the basis of spectral data?
Line 176 what is the effective spectra?
Line 179 - What information do these layers include?
References in table 2 - reference should be given as a number in [] referring to the number of this publication in References list at the end of the article
Lines 288-289, 292-293 should be a part of Methods (particulary subsection 2.4 )
Lines 323-324 Correlation cannot be measured in %. It is estimated using coeffitient of correlation (Pearson, Spearman, etc.). It is unitless and varies strictly between -1 and 1.
Table 6 - the content shifted
Figure 6 -how were soil texture classes determined?
Conclusion should be more specific and include some quantitative data (like RMSE, R2, etc.)
Author Response
Authors’ Response to the Reviewer Comments
Journal: Soil Systems- MDPI
Manuscript ID: 528602
Title of MS: Spatial prediction of soil texture fractions in arid regions of Iran
Authors: Elham Mehrabi-Gohari1, Hamidreza Matinfar2,*,Azam Jafari3, Ruhollah Taghizadeh-Mehrjardi4,5, John Triantafilis6
The authors would like to thank the Editor and the reviewers for their precious time and valuable comments. We have carefully addressed all the comments. The corresponding changes and refinements made in the revised paper are summarized in our response below.
Comments from the editors and reviewers:
The reference in Introduction are still in the wrong order. I suggest you properly check it. References should be numbered in the text in order of appearance. Please, refer to https://www.mdpi.com/journal/soilsystems/instructions.
Answer: Thank you for the comment. This edit was made.
Table 2 (line 112) is mentioned before table 1 (line 189).
Answer: Thank you for the comment. This edit was made.
Tables should be placed after their first mention in the text.
Answer: Thank you for the comment. This edit was made.
The caption for Figure 3, 4 are incorrect. It should look like
Answer: Thank you for the comment. This edit was made.
Figure 3 Spectra of 10 selected soil sample: a) reflectance, b)......
Answer: Thank you for the comment. This edit was made.
And Figure 4 shows the importance of the variables, so it should be named properly.
Answer: Thank you for the comment. This edit was made.
Line 34 - MrVBF should be explained
Answer: Thank you for the comment. This edit was made.
Line 46 - such as ; bulk density (; should be removed)
Answer: Thank you for the comment. This edit was made.
Lines 58, 59 - have been used, have been used
Answer: Thank you for the comment. This edit was made.
Lines 169-170 modelling was done on the basis of spectral data?
Answer: yes. At first spectroscopic data were pre-processed using the first derivative, the second derivative with the middle filter and the Savitzky & Golay filter and then the plsr method was used for modeling and the best spectra for clay and sand and silt estimation were separated by weighting.
Line 176 what is the effective spectra?
Answer: 879,890 for clay; 852,879 for sand; 1456,1998 for silt.
Line 179 - What information do these layers include?
Answer: Thank you for the comment, The spectral layer contains reflections of electromagnetic waves in the spectral range of 400–2500 nm from the soil particles, these data are the result of the collision of the waves with the soil constituents, which vary depending on the composition of the reflections and the absorption, such as water molecules of wavelengths 1400 and 1900nm. They are absorbed so the presence of clay increases the absorption. On the other hand, the particle size in the reflection is such that the fine particles like clay have more reflection than coarse particles like sand.
References in table 2 - reference should be given as a number in [] referring to the number of this publication in References list at the end of the article
Answer: Thank you for the comment. This edit was made.
Lines 288-289, 292-293 should be a part of Methods (particulary subsection 2.4 )
Answer: Thank you for the comment. In the Materials and Methods section, it is stated that these parameters are entered as auxiliary data. Here the effect of each of them on the estimation of clay and sand and silt factors is compared .
Lines 323-324 Correlation cannot be measured in %. It is estimated using coeffitient of correlation (Pearson, Spearman, etc.). It is unitless and varies strictly between -1 and 1.
Answer: Thank you for the comment. This edit was made.
Table 6 - the content shifted
Answer: Thank you for the comment. This edit was made.
Figure 6 -how were soil texture classes determined?
Answer: At first, separate clay, sand, and silt maps were prepared in the GIS software. Then, in the SAGA software, three maps were combined and the final map was prepared.
Conclusion should be more specific and include some quantitative data (like RMSE, R2, etc.)
Answer: Thank you for the comment. This edit was made and parts added.
Author Response File: Author Response.doc
