UAV and LiDAR Data in the Service of Bank Gully Erosion Measurement in Rambla de Algeciras Lakeshore

Round 1

Reviewer 1 Report

water-900223

This paper is a revised manuscript of a previous submitted article and addresses an interesting and up-to-date subject, although a bit medley. The manuscript should be improved in some places, but it adheres to Water journal policies and topics. Below I have inserted some recommendations through which the scientific value and readability of the manuscript could be improved.

            The structure of the manuscript need to be improved, and I strongly recommend a classic structure of Introduction, M&M, Results, Discussions, Conclusions. For such paper a separated Discussion chapter is mandatory. Also, you include in Results “Data acquisition” and “Data processing”, those belong to M&M!

Figure 8, 9 and 10 are the most important in the manuscript, and should have the same colour scheme for erosion and accumulation. Fig 8 and 9 are the same (white for deposit), then Fig 10 has a different symbology (white for erosion). You should correct this aspect, maybe keep the symbology from Fig 10, it is more appropriate. Also, the “Legend” from each figure must be improved, write what that intervals represents and the measurement unit.

In my opinion the manuscript is inhomogeneous because of the combination of UAV SfM and Lidar. You could have done 2 articles with the data from this manuscript, one regarding the 3 case studies with UAV technology, and one with Lidar. Nevertheless, the work is genuine and presents a good and interesting case study.

Author Response

Point 1: The structure of the manuscript need to be improved, and I strongly recommend a classic structure of Introduction, M&M, Results, Discussions, Conclusions. For such paper a separated Discussion chapter is mandatory. Also, you include in Results “Data acquisition” and “Data processing”, those belong to M&M!

Response 1: We have created a discussion section. In this section, we have discussed the methodology and the results with other scientific works.

Point 2: Figure 8, 9 and 10 are the most important in the manuscript, and should have the same colour scheme for erosion and accumulation. Fig 8 and 9 are the same (white for deposit), then Fig 10 has a different symbology (white for erosion). You should correct this aspect, maybe keep the symbology from Fig 10, it is more appropriate. Also, the “Legend” from each figure must be improved, write what that intervals represents and the measurement unit.

Response 2: The figures (10,11, 12 and 15) have the same legend (red = erosion and blue = accumulation)..

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors submitted a revised manuscript dealing with a very interesting theme of topography survey with UAV. Authors present a now-familiar process of UAV data processing. Unfortunately, from previous submission authors made only small entry into the article and changed a manuscript very little. Exactly, authors added some lines to the end of the introduction, changed the order of the chapters/sub-chapters and reduce the general description in the methods section. For that, the main problems of the presented paper still preserving. Structure of the paper was slightly improved but in the methods section are much general irrelevant information and on the other hands’ description of error estimation (principally FIS model) is very insufficient. In the Result section, chapter 3.1, 3.2.1., 3.2.3 describe data acquisition and processing and should be transferred to the methods section.

Main comments:

1)  The main weakness of this study is ambiguity between the study focused on the technology and case study describing the process of bank gully erosion. The manuscript is located on the halfway between them. Methods section bring a large amount of general information and statement, but on the other side, there still missing substantial information about point cloud generation, FIS model description (inputs, outputs and rules definition).

2) In the manuscript missing the main idea for error estimation. In the paper is a mix of data models (raster DEM, point clouds), data types (UAV, LiDAR) and errors estimation (cross-validation, C2C, FIS, LOD, coregistration or georeferencing errors). But missing main information: What error do they want to identify? Is the main aim to identify deviation errors for minimum DoD identification, or specify detection limits, or only generate general error information? According to this question, specific types of errors are selected. From the presented papers I do not understand why authors used cross-validation and raster creation when C2C/ M3C2 compute differences from point clouds and what is final differences in volume estimation from these 3 approaches? In the FIS model missing exact information about inputs, outputs, rules definition and process of slope and roughness identification as well as final elevation error interpretation in different morphological zones of gullies. From a process perspective, it is essential to evaluate deviation errors resulting from DTM processing as estimated by critical evaluation of uncertainties arising from measurement errors, control points accuracy, coregistration, and data coordinate transformation errors. Therefore, before interpreting the results of surface differences, it is necessary to express the measurement errors, as well as the errors connected with the registration process and calculation error propagation.  

3) In the second part of the MS missing statistical evaluation of the relationship between the physical properties of environments (precipitation, temperature, water level fluctuation) with gully erosion.

Author Response

Point 2:

In the manuscript missing the main idea for error estimation. In the paper is a mix of data models (raster DEM, point clouds), data types (UAV, LiDAR) and errors estimation (cross-validation, C2C, FIS, LOD, coregistration or georeferencing errors). But missing main information: What error do they want to

identify? Is the main aim to identify deviation errors for minimum DoD identification, or specify detection

limits, or only generate general error information? According to this question, specific types of errors are

selected. From the presented papers I do not understand why authors used cross-validation and raster

creation when C2C/ M3C2 compute differences from point clouds and what is final differences in volume

estimation from these 3 approaches? In the FIS model missing exact information about inputs, outputs,

rules definition and process of slope and roughness identification as well as final elevation error

interpretation in different morphological zones of gullies. From a process perspective, it is essential to

evaluate deviation errors resulting from DTM processing as estimated by critical evaluation of

uncertainties arising from measurement errors, control points accuracy, coregistration, and data

coordinate transformation errors. Therefore, before interpreting the results of surface differences, it is

necessary to express the measurement errors, as well as the errors connected with the registration

process and calculation error propagation.

Reponse 2:

- We eliminated the C2C technique and focused just on the FIS and Cross-validation method.

- Our objective is the estimation with high accuracy of the SfM-MVS products( DTM and

orthophotos), so we first estimated the accuracy of the SFM-MVS point clouds using the

cross-validation method before building an error model using the FIS model.

- from L574 to L583, In the discussion section, we compared the accuracy of our data with

other works

- The M3C2 algorithm is used to estimate the volume eroded on the B.G-3 gully.

- Forme L 225 – L281, In this pragraph we explained the input, output data and the rules of

the FIS model.

Point 3 : In the second part of the MS missing statistical evaluation of the relationship between the

physical properties of environments (precipitation, temperature, water level fluctuation) with gully

erosion.

Reponse 3: I agree with you, the correlation analysis between the eroded volume and the

physical properties of environments can explain the gully initiation factors, but this

correlation can be the subject of another article and unfortunately, we cannot put all the

results in this article.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors have done a lot of work to study erosion dynamics. However, the adequacy of the obtained results is highly questionable due to the lack of description of a number of methodological steps.

Rows 23-26: Not a complete definition of erosion, as erosion and runoff formation on the Earth's land, occurs not only as a result of rainwater but also melt runoff.

In our opinion, in the Introduction too much attention is paid to theses on the concept and role of erosion in the formation of relief and soil degradation. This is understandable for specialists. The text talks about the dangers of soil erosion. What for? After all, the purpose of the work is to study gullies, that is, linear forms of erosion. And the development of gullies, in contrast to soil erosion, is very local and discrete in geographical space.

Rows 51-53: In the list of instrumental methods, the method of monitoring gullies by ground-based laser scanning now occupies an important place. Please like to reflect on this fact in the text.

Rows 103-109: The choice of this gully for assessing its contribution to the siltation of the reservoir is somewhat doubtful. It is not located near the coastal zone of the reservoir and is 250 m away from it.

The morphology, types of processes that form the slopes, and the rate of linear growth of gullies are largely determined by lithology. In the description of test gullies, nothing is said about the lithological composition of the rocks in which they were formed. Is the rock mass homogeneous? What is the catchment area of ​​these gullies? What is the total density of the gullies dissection of the reservoir catchment?

Photographs of test gullies can be included for a better understanding of the study sites.

Aerial photogrammetric works have clear instructions, following which is not described by the authors. What software did the authors use for UAV flight? What is the longitudinal and transverse overlapping of images during survey and at what altitude? What is the accuracy of reference point coordinates determination (a detailed table is needed)? In what mode was GCP surveyed? Static? Kinematics? Answers to these questions should be included in the appropriate section of the manuscript.

The need to implement the procedures described in section 2.4.1. is not clear. Errors of what exactly do the authors describe? Triangulator modeling? If so, why wasn't interpolation compared with other methods? What is the density of the initial dense cloud and why did the authors decide to choose the 5 cm resolution? Why wasn't a similar procedure done to evaluate the Lidar data interpolation?

Rows 368-369 do not make any scientific sense, as it is obvious that the greater the resolution of the model, the greater the roughness it will describe.

Rows 387-390. "The points installed at the bottom of the gullies have a low accuracy compared to the others. They don't exceed 4 cm at best. This can be explained by the poor reception of the GNSS signal at the bottom of the gullies." The manuscript does not describe the procedure for post-processing GNSS data. Has it been done? At what distance was the base station located?

Table 2. Correlation between the camera parameters - The values of the parameters themselves are not presented. It is also not clear from the text, the correlation of parameters with what the authors analyze?

Figures 10-12. A - it is necessary to bring the color scale of difference in an adequate form - erosion - red shades, accumulation - blue, areas with zero dynamics - white. The scale step is 1 m, the area of zero changes with a step equal to the accuracy of determining of coordinates.

Correlation analysis is required between the obtained results and various DEM based products - slope map, TWI, etc. So far, visual analyzing of Figures 10-12, it seems that C2C minimized the error, but added planar and height distortions in dx, dy and dz (check the "stable" areas - roads, areas with zero slopes, etc. - in them the difference in height should be minimal, for roads at all zero).

Figure 15 should also lead to a clear color scale, as well as make a few cut-maps with an enlarged presentation of areas of change.

There is no cause-effect analysis in the section describing the results. As well as any other analysis of the results, not just a description of them.

Rows 757-759 - ????

All of the above leads to the fact that the current iteration of the manuscript cannot be recommended for publication, and it may be a reconsider after major revision (control missing in some experiments).

"The main goal of this paper is to demonstrate whether the SfM-MVS-4D model generated from photos taken by a UAV and the LiDAR PNOA data, are able to accurately quantify the sediment budget on the banks gullies of the lakeshores. The answer is yes ". The good quality of the research carried out, the well-founded use of modern instrumental methods for studying gully erosion allows us to conclude: the authors have every reason to draw this conclusion, but after the article has been revised.

Author Response

Point 1: Rows 23-26: Not a complete definition of erosion, as erosion and runoff formation

on the Earth's land, occurs not only as a result of rainwater but also melt runoff.

Point 2 : In our opinion, in the Introduction too much attention is paid to theses on the

concept and role of erosion in the formation of relief and soil degradation. This is

understandable for specialists. The text talks about the dangers of soil erosion. What for?

After all, the purpose of the work is to study gullies, that is, linear forms of erosion. And the development of gullies, in contrast to soil erosion, is very local and discrete in geographical

space.

Response 1 and 2: We have modified the introduction, we have given a more precise

definition to gullying and bank gullying, instead of going back to the general definition of

water erosion, then we have explained the effects of bank gullying.

Point 3: Rows 51-53: In the list of instrumental methods, the method of monitoring gullies by ground-based laser scanning now occupies an important place. Please like to reflect on this fact in the text.

Reponse 3: The LiDAR methodology was briefly described because we did not carry out

LiDAR data collection missions, so we just processing the raw data.

Point 4: The choice of this gully for assessing its contribution to the siltation of the reservoir is

somewhat doubtful. It is not located near the coastal zone of the reservoir and is 250 m away from it.

Response 4: In the description of the bank gully B.G-1 we gave the reason for the choice of

this site.

Point 5: The morphology, types of processes that form the slopes, and the rate of linear growth of

gullies are largely determined by lithology. In the description of test gullies, nothing is said about the

lithological composition of the rocks in which they were formed. Is the rock mass homogeneous?

What is the catchment area of these gullies? What is the total density of the gullies dissection of the

reservoir catchment?

Reponse 5:

- Frome L99 to L103 : A description of the lithology of the study area.

- Figure 16 : Gully density and L 189 to L191in this paragraph we compared the gully

density with other studies

Point 6 : Photographs of test gullies can be included for a better understanding of the study sites.

Reponse 6 : we have included in the text a photographs of test bank gullies (Figure 2)

Point 7 : Aerial photogrammetric works have clear instructions, following which is not described by the

authors. What software did the authors use for UAV flight? What is the longitudinal and transverse

overlapping of images during survey and at what altitude? What is the accuracy of reference point

coordinates determination (a detailed table is needed)? In what mode was GCP surveyed? Static?

Kinematics? Answers to these questions should be included in the appropriate section of the

manuscript.

Reponse 7 :

- From L 175 to L 177 In this pragraph, we mentioned the use of Pix4dcapture software for

UAV flight.

- From L181 to L183, we indicated Longitudinal and transverse overlapping, In table 1

“Characteristics of UAV photo missions” we gave flight altitude for each mission.

- We agree with you, but to keep it simple we have given Table 3 which summarizes the

uncertainty of these points, as explained in several studies 1

https://0-www-mdpi-com.brum.beds.ac.uk/2072-4292/4/11/3390/html

https://0-link-springer-com.brum.beds.ac.uk/article/10.1007/s12665-018-7817-4

- L 189 to L191: In this pragraph, we explained GPS mode to collect GCP points.

Point 8: The need to implement the procedures described in section 2.4.1. is not clear. Errors of what

exactly do the authors describe? Triangulator modeling? If so, why wasn't interpolation compared with

other methods? What is the density of the initial dense cloud and why did the authors decide to

choose the 5 cm resolution? Why wasn't a similar procedure done to evaluate the Lidar data

interpolation?

Reponse 8:

- In this part, our objective is the estimation with high accuracy of the SfM-MVS

products( DTM and orthophotos), so we first estimated the accuracy of the SFM-MVS point

clouds using the cross-validation method before building an error model using the FIS model.

- In our case, the precision of 5 cm is the optimal size that minimizes the loss of information

on the topographic complexity of the field.

Point 9: Table 2. Correlation between the camera parameters - The values of the parameters

themselves are not presented. It is also not clear from the text, the correlation of parameters with what

the authors analyze?

Reponse 9: L550 to L557 In this pragraph, we explained the usefulness of the correlation

table between camera parameters. Table 4 confirms the choice of acquisition configuration.

Point 10: Figures 10-12. A - it is necessary to bring the color scale of difference in an adequate form -

erosion - red shades, accumulation - blue, areas with zero dynamics - white. The scale step is 1 m,

the area of zero changes with a step equal to the accuracy of determining of coordinates.

Reponse 10 : The figures (10,11, 12 and 15) have the same legend (red = erosion and blue =

accumulation).

Point 11: Correlation analysis is required between the obtained results and various DEM based products – slope map, TWI, etc. So far, visual analyzing of Figures 10-12, it seems that C2C minimized the error, but added planar and height distortions in dx, dy and dz (check the "stable" areas - roads, areas with zero slopes, etc. – in them the difference in height should be minimal, for roads at all zero).

Reponse 11 :

I agree with you, the correlation analysis between the eroded volume and the topography can explain the gully initiation factors, but this correlation can be the subject of another article and unfortunately, we cannot put all the results in this article.

Point 12: There is no cause-effect analysis in the section describing the results. As well as any other analysis of the results, not just a description of them.

Reponse 12: In the results section, we presented only the results, but in the discussion section, we compared our results with other works.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Water-900223-peer-review-v2        

The revised manuscript demonstrates the author’s commitment in improving the overall paper, thus obtaining a cohesive and interesting article.

• The overall length of the manuscript was increased with 2 pages, that include useful additional information and citations in Introduction and M&M, as well as some very important figures such as Fig 2 and 16
• Figures were improved as requested in the review eg. Fig 11-13 and 15
• The manuscript was restructured and now include a large chapter of Discussion with subchapters

Reviewer 2 Report

The manuscript was substantially improved. For the publication I recommend to improve the quality of same figure, mainly Fig. 6 (graphic quality) and Fig. 10, 11 and 12 (size of parts D and E is unreadable).

Reviewer 3 Report

The authors have made a major revision of the text of the paper. The additions made to the article and the explanations of the authors significantly improved the understanding of the research. In my opinion, the paper can be published.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

water-816254

This paper addresses an interesting and up-to-date subject. The manuscript should be improved in some places, but it adheres to Water journal policies and topics. Below I have inserted some recommendations through which the scientific value and readability of the manuscript could be improved.

            Minor/major comments for the authors

I appreciate the genuine work carried out throughout the research, but some M&M aspects are in too great detail, and some can consider this redundant or unnecessary at this scientific level (eg. photogrammetry workflow, error modelling). In my opinion you can “trim” the manuscript of 2-3 pages, because at 26 pages as it is, it is a little hard to read. (this is my opinion)

At UAV Data Processing subchapter, you can also add the mission planner that you used. Also, in the future I would recommend you use the same “job/mission plan” for monitoring land/gullies. The correct way is to have the same height, the same number or images and the same GCPs (if possible) between missions. In that way, the changes/displacements are correctly monitored.

Also, when you mention the data processing step, you wrote “this processing step generally takes a few days” and “post-processing of SfM-MVS data sets requires a working time ranging from a few weeks to several months”. It is true that this step can take a lot of time, but what you wrote is the worst-case scenario, when you have a low-grade processing power PC. In this kind of studies, graphic stations are used and the period is severely shortened. In my opinion you should rewrite this idea and remove this exaggerated remark.

R225 “according to the type and type of type of ….” Please correct

Figure 6 has very bad resolution. Please try to improve it

In many places throughout the manuscript you say “GCP points”. This is a pleonasm, so please correct it, you can just say GCPs

Also, in many places you say “GPS surveys”, the correct way is DGPS or GNSS, not only GPS.

Figure 7 has to be improved (from Erreur to English and only 3 decimals after comma)

Figure 8, 9 and 10 are the most important in the manuscript, and should have the same colour scheme for erosion and accumulation. Fig 8 and 9 are the same, then Fig 10 has a different symbology. You should correct this aspect, maybe keep the symbology from Fig 10, it is more appropriate. Also, the “Legend” from each figure must be improved, write what that intervals represents and the measurement unit.

Finally, I would like to congratulate the authors for the effort and scope of the article. It presents an interesting topic and has good readability. Regarding the manuscript, the current form needs medium revisions.

Reviewer 2 Report

Thank you very much for submitting your manuscript to "water". I appreciate that you have made substantial effort to develop the ideas for this manuscript and conduct the research. The research is interesting but it is difficult to assess at this stage whether it merits publication. This is mainly because the aim and objectives of your work are not clearly stated. It is therefore difficult to ascertain what you are trying to accomplish with this paper. My understanding is that you are trying to identify temporal change in gullies and you are comparing the point clouds to be able to do so. If this is the case, I suggest you expand the introduction to capture the aim and objective of your work and you summarise the methodology coherently to make it succint, sound and sharp (the same applies to the results) and then resubmit. I have rejected the paper so that you have sufficient time to address all these changes at your own pace. They are substantial and will require a re-write of the manuscript.

Reviewer 3 Report

The paper seems to bring some very interesting contribution to the landscape mapping by using UAV technology. However, the structure of the paper results to be rather confusing so that the true added value does not fully emerge from the work conducted and described in detail. 
Authors introduce the topic starting from the wider explanation of soil erosion and general statement (see L28 – 36). This distracts the attention from the specific UAV description in the field of erosion identification and finally a comprehensive review of UAV used in soil erosion is missing. On the other side, the Discussion section is merged with the result section without discussion with literature. The main weakness of this study is ambiguity between study focused on the technology and the case study describing process of bank gully erosion. The manuscript is located on the halfway between them. Methods section bring a large amount of general information and unnecessary statement (se L131 – 170), but on the other side there still missing substantial information about point cloud generation. Section 2.3.1 Error modelling need more work for a better description of data layer processing. Section Results and Discussion is a combination of methods, results and discussion. From L286 to L402 authors presented methods and should be transferred to this section. I suggest to modify paper in one way: but to improving and better description former Data error estimation section and comparison with independent dataset (LiDAR, check point) or focus to morphological process of shoreline erosion with comprehensive literature review and discussion in this climate environment.

Important note: Cross validation express accuracy of resampling/interpolation and do not pointed to the point cloud quality. But there is missing information about point cloud density changes during this process.  Please, distinguish differences between data resampling, smoothing and averaging.