Mapping of Rill Erosion of the Middle Volga (Russia) Region Using Deep Neural Network

Round 1

Reviewer 1 Report

The paper entitled "Mapping of rill erosion of the Middle Volga (Russia) using artificial intelligence", is contributing with good data and the paper could be a useful contribution to ISPRS International Journal of Geo-Information. However, the authors need explain with more details the Materials and methods and improve the results and discussion.

 

I have a variety of remarks that I would recommend the author to consider, as listed below:

 

Tittle: The title itself is misleading and does not reflect the aims of the work. Please, consider changing the tittle to “Mapping of rill erosion of the Middle Volga (Russia) using artificial intelligence and Sentinel-2 data”.

 

In Abstract: what are the main results? What is the main conclusion that the study reached?

 

In Introduction: The context of this work has been poorly defined, and motivation is greatly lacking. While the manuscript presents a good case study and methodology, the overall novelty is low. Try to further explain the importance/novelty of your research as opposed to previous studies.

 

In Introduction: More citations in some statements are needed.

 

In Introduction: The objective of this study is not clear.

 

In Introduction: No information is provided about new methodology for large-scale automated mapping of erosion processes, namely stream erosion, based on Sentinel-2 data. Please clarify.

 

In Introduction: No information is provided about GAM models. Please clarify.

 

Line 48: To check this citation.

 

Line 49: To inset more references.

 

Lines 50-51: Need reference.

 

Line 56: “Different approaches are used for this.” Please add more information and clarify about these approaches.

 

Lines 57-58: Need reference.

 

Lines 76-78: I recommend citing more references.

 

Lines 79-80: Missed mentioning the SWAT model that is one more cited models around the world currently.

 

Line 120: To insert coordinates of the study area.

 

Lines 137-140: Need references.

 

Lines 142-144: Need reference.

 

Line 154: Improve the quality io Figure 1.

 

In material and Methods: The methodology and data are poorly explained. Materials and methods must be improved.

 

Lines 158-167: More information about this information are necessary.

 

Lines 171-173. More information about GEE technique must be presented in manuscript.

 

Line 184: How the ground truth were obtained? Please to insert more information.

 

Lines 247-248: “For most of the selected factors, no strong correlations were found with the total length of rill erosion.” The information is poorly explained. Please quantify this information.

 

Lines 264-266. “Apparently, with increasing temperature and climate change due to global warming, the character of precipitation changes - they may fall rarely, but with great intensity.” This statement is not supported in this study. Please, re-written this sentence.

 

Line 271: Table 1 must be moved to Material and Methods.

 

Line 271: How Table 1 factors were obtained (source, accuracy of data)? Is not clear.

 

Lines 282-283: This sentence must be moved to Material and Methods. In addition, this information must be clarified.

 

Line 296: How precipitation data were obtained? All data exposed in item Results must be presented in Material and Methods (source, resolution and accuracy of data).

 

Line 381: Soil maps of study area is not presented in supplementary material. Furthermore, how was performed relationship between length of the erosion network on the prevailing soil type?

 

Line 455: “MANOVA” is not presented in methodology.

 

Improve Figures 11, 12, 13, 14 and 15: X and Y axis not presented.

 

The model is poorly explained. I recommend creating a flowchart with methodology used in this study.

 

How length of ephemeral gully erosion in the study were classified and determined?

How normalized distribution is calculated?

 

A discussion chapter would be beneficial. I recommend separate Results and Discussion.

 

Conclusion chapter should be improved, as well as elaborating future directions for the research.

Author Response

Dear Reviewer. We have tried to consider all comments to improve the manuscript. In the attachment there is a point-by-point answer to the given questions

Author Response File: Author Response.docx

Reviewer 2 Report

The author presents an interesting and relevant research combining machine learning, statistical analysis and geography. The article is well written, and it is easy to follow the research structure, its methodology and results. The English spelling has some style and grammar problems, especially in sections 1-3, but nowhere it makes the text unclear. I suggest some minor adaptation to the manuscript before publications, mainly related to the description of the novel methodology to the identify the areas with linear rill erosion using a deep neural network.

Comments for author File: Comments.pdf

Author Response

Dear Reviewer. We have tried to consider all comments to improve the manuscript. In the attachment there is a point-by-point answer to the given questions

Author Response File: Author Response.docx

Reviewer 3 Report

This paper takes the problem of the land erosion in the large scale and geo-hazards management. The proposed methodology with combined remote sensing tools and neural networks is novel approach. The paper structure and text preparation are not appropriate and  the some sections need correction and supplementation. Below I mention some problems with your article.

 

General comments:

1. Paper’s title is not to accept. The article describes the gullies phenomena in the large spatial scale not rill erosion forms, which needs local scale and more detailed resolution. I propose change the title to: “Mapping of gully erosion of the Middle Volga (Russia) region using remote sensing and artificial intelligence tools’.
2. Abstract needs correction, please to reduce this text to: aims, methods and obtained results.
3. Introduction section needs correction. Please indicate reasons of your investigation and aims of this study. I propose only descripting of the problems of the remote sensing methods using to detection and monitoring of gully erosion processes..
4. Study area need supplementation with information of natural environment, e.g. geological sediments, relief and landforms. I propose to add too information about agriculture characteristic, e.g. cultivation, tillage technics and so one.
5. Material and Methods section need supplementation. I propose to add one paragraph about the classification of the erosional forms, such as: valleys, channels, gullies and others. How you calculate the: ’rill erosion density’? You can give some information about correlation analysis too.
6. Your obtained results can be compare with subject literature. At present in the text is lack of discussion.
7. Conclusion needs supplementation according to the obtain results.

More detailed comments are provided in the text (enclosed pdf).

 

To discussion:

1. Problem with geomorphic terminology, you use in the text both term: ‘rill erosion’ and ‘gully erosion’. In the theory of geomorphology there are different processes see:
2. Bryan R.B. (ed.), 1987. Rill erosion. Processes and significance. Catena Supplement 8, Braunschwieg.
3. Poesen J., Nachtergaele J., Verstraeten G., Valentin C., 2003. Gully erosion and environ­mental change: importance and research needs. Catena 50 (2-4): 91–133.
4. Ionita I., 2006. Gully development in the Moldavian Plateau of Romania. Catena 68:133– 140.

And they need other spatial scale, see:

Brunsden, D., 1996. Geomorphological events and landform change. Zeitschrift für Geomor­phologie 40(3): 273–288.

Ionita I., 2006. Gully development in the Moldavian Plateau of Romania. Catena 68:133– 140.

Janicki G., 2014: Transformation of upland wash slope-a case study from the Lublin Upland (SE Poland), Annales Universitatis Mariae Curie-Sklodowska 69 (1), 31-48

 

2. You use the different terms which have not clear definition, e,g, soil erosion, stream erosion, rill erosion and gully erosion. And they are not similar terms.
3. In my opinion you can to explorer the ephemeral gullies theory. They are erosional forms between channel erosion and gully erosion form (see: Poesen J., Nachtergaele J., Verstraeten G., Valentin C., 2003; Janicki 2014, 2016).

Comments for author File: Comments.pdf

Author Response

Linear erosion processes on arable land are complex and interrelated. The following classification of erosion types in terms of increasing damage to the soil cover is widely applicable: areas of raindrop erosion - as a rule, are confined to the watershed parts of the slope with about zero gradients. Erosion, in this case, has the mechanism of a detachment of the lightest and weakly consolidated particles from the impact of raindrops. The more intense the rainfall, the greater the damage. These areas do not fall into the linear type of erosion. As the gradient increases and soils become saturated with moisture, areas of micro-rill erosion appear. This area can already be classified as a linear erosion area. Field methods can only assess such areas. The width of washouts reaches 15-20 cm. The concentration of micro-rill as they move down the slope leads to areas of rill erosion - washouts here reach up to 2 m, passing into areas with ephemeral gullies reaching up to 10 m. These erosion forms will become permanent gully areas without special erosion control measures.

There is a vague boundary by which permanent gullies differ from ephemeral gullies. When discussing with foreign colleagues-geomorphologists, we found out that most of them distinguish permanent gullies from ephemeral ones by the criterion of ploughability - if a gully is ploughable, then it is ephemeral. In the literature of the CIS countries, other criteria of distinction are given. But there is no unanimity in this question either. In Kazan geomorphological scholarship, the following criteria are used: the presence of a distinct headcut; absence of sodding on the gully sides; a developed out own profile of the gully which differs from the profile of the slope, outcrops of bedrock. Ephemeral gullies and rill erosion are differentiated differently, but one way or another, classification in mapping is, in our opinion, only possible by fieldwork or by ultra-high-resolution DEM data. Distinguishing ephemeral gullies from permanent ones by satellite imagery data is possible.

In our work, we prepared the training data so that the neural network had no uncertainty in distinguishing ephemeral gullies from permanent gullies. The situation with soil washouts is more complicated, and in our study, we did not differentiate between ephemeral gullies and rill erosion. We distinguished all the washouts visible on the satellite image but are not permanent gullies. Considering the image's resolution, we can confidently say that rill erosion is detected successfully, even though their width reaches 5 m only in extremely rare cases. However, rills are mapped by indirect signs, i.e., change of phototone to lighter side as the rill thalweg is approached.

We understand that in preparing the manuscript, we allowed a situation of uncertainty as to what we recognized since the terms "rills," "ephemeral gully," etc., are given in the text, although they mean the same thing in our case. We corrected the text and put the terms together, and an appropriate disclaimer was made in the introduction.

In terms of technical comments in the PDF, we have tried to take into account all the reviewer's advice and thanked him for his careful reading of the manuscript.