Next Article in Journal
An Efficient Calibration System of Optical Interferometer for Measuring Middle and Upper Atmospheric Wind
Next Article in Special Issue
Potential for Peatland Water Table Depth Monitoring Using Sentinel-1 SAR Backscatter: Case Study of Forsinard Flows, Scotland, UK
Previous Article in Journal
Contribution to Uncertainty Propagation Associated with On-Site Calibration of Infrasound Monitoring Systems
Previous Article in Special Issue
Cloud-Based Remote Sensing for Wetland Monitoring—A Review
 
 
Article
Peer-Review Record

The Spatial Analysis of Vegetation Cover and Permafrost Degradation for a Subarctic Palsa Mire Based on UAS Photogrammetry and GPR Data in the Kola Peninsula

Remote Sens. 2023, 15(7), 1896; https://doi.org/10.3390/rs15071896
by Natalya Krutskikh 1, Pavel Ryazantsev 2,*, Pavel Ignashov 3 and Alexey Kabonen 4
Reviewer 1: Anonymous
Reviewer 2:
Reviewer 3:
Remote Sens. 2023, 15(7), 1896; https://doi.org/10.3390/rs15071896
Submission received: 15 February 2023 / Revised: 24 March 2023 / Accepted: 30 March 2023 / Published: 31 March 2023
(This article belongs to the Special Issue Application of Remote Sensing for Monitoring of Peatlands)

Round 1

Reviewer 1 Report

Dear authors,

Firstly, I would like to express my gratitude to the authors for their manuscript; it was a pleasure to read. The paper presents a novel approach to understanding the relationship between permafrost degradation in palsas mires and surface traits, such as landform and vegetation cover, through the combination of aerial photogrammetry, in situ vegetation surveys, and Ground Penetrating Radar (GPR) surveys. However, I encourage the authors to put in additional work to further improve the quality of the manuscript.

General comments

In the introduction, the authors could emphasize the significance of the study and the potential implications of failing to monitor permafrost. While the citations in the introduction are appropriate, more recent publications should be included in some areas.

The methodology used in the study needs to be explained thoroughly to ensure reproducibility. The authors should clearly justify each methodology used, including more details (see specific comments). For example, it needs to be explained and justified the resampling of pixels and the method used for that. More details are needed in the land cover classification methodology. Why did the authors choose these three methodologies instead of just one and which are they using for their results. Also it is very important to explain how the training areas were created and their source. Was it visual interpretation? Was it data collected on the field? Please make sure to include more details.

The results are well presented, but some data could be better integrated into the discussion. I would like to recommend to do two more plots that I think can make the paper even more interesting:

1)    Add an additional figure (figure 5b) that includes all pixels, not just the training data

2)    It would be really interesting to see GPR signal characteristics plot by each LC

The authors should discuss the uncertainties of the land cover classification and how it affected the correlation with the GPR, as well as potential improvements to the classification.

In the results, it is unclear how did you use the manual probing data collected in the field. When did you use this data? Should be explained more thoroughly

Finally, the references are appropriate, but some citations could be repeated in some paragraphs when appropriate.

Specific comments

Page 1 Paragraph 1: More recent references should be added into the first paragraph when talking about how climate change is affecting palsa mires and permafrost areas.

Page 1 Paragraph 2 - line 3: Add a better description of palsa mires. See for example: Ballantyne C. K., 2018; Tarnocai et al., 2009

Page 2 – Paragraph 2 – line 4: Add references about the following sentence: “The reason is the higher rates of palsa decay and the associated significant changes, so that individual trends of ecosystem change can be detected through observations of short duration”

Page 2: Difficult to read and follow the following sentence, please re-write:

This conclusion cannot be reliable, however, since the assessment was based only on the mean annual temperature and thickness of the active layer on palsas. However palsa mires are known to undergo substantial lateral permafrost degradation [5], while in the vertical projection, according to a model suggested by Seppälä [12], demonstrates seasonal growth.“

Page 2 – Paragraph 2: More references could be added in the end of the paragraph

Page 2: Add also something about the inaccessibility of the areas when talking about the advantages of remote sensing

Page 2 – Paragraph 3 - line 3: Add more references, for example:     

- https://doi.org/10.5194/bg-2023-17

- Fewster etal 2022

 Page 2 – Paragraph 3 - line 3: Which study are you talking about, make it more clear.

useful tool in the study and modelling…”

 Page 2 – More references should be added here

“and subarctic regions are widely used in vegetation assessments [27, 29]”

 Figure 1: Map elements missing:

-       North arrow.

-       Scale bar in the panel b.  

-       It might be useful to add a), b) and c) to explain the image

-       Units in the legend

-       Reference to the elevation data (is that from the flights?).

-       Date of satellite imagery and band combination is missing

Page 4: At the beginning of page 4, Is this now the methodology? Add a subtitle if that is the case. I can see that below starts the methodology. However, the style of this paragraph sounds like methodology. Please make sure to either make it clear is part of the site description or methodology

Page 4, methods: There are plenty of studies for that. Need to add relevant references. E.g. 31 and 32 “Unmanned aerial systems (UAS) have been successfully used in vegetation monitoring [40].”

Page 4: I think it is quite expensive to fly drones

“key benefits are the relatively inexpensive maintenance and high-resolution”

Page 4: Add reference number 28 as well

“…grid cell resolution and explore their attributes [41].”

Page 5: Is it possible to see this in an image?

“As the mire is rather large and battery capacity (25 min at maximum) limits the UAS flight time, the territory to be covered by the aerial survey was divided into two adjacent 350 × 300 m plots using Pix4DCapture flight planning application.”

Page 5 – Paragraph 1: Did you use ground control points with a differential GPS? Need to talk about this. If not why they were not necessary and which are the implications

Page 5: You introduce machine learning but it doesn’t say what will it be for. If you will introduce it here, make sure you say for what. You could reference the section you’ll later talk about it.

The photogrammetry processing results specified in Table 1 formed the input for machine learning”

Page 5: Specify the methodology of how did you reduce the resolution. The original resolution was 1.96 cm and then you aggregate pixels to get 50cm? It needs more justification or explain which are the implications of doing this.

“To enable local-scale land cover classification, the orthophoto and DEM spatial resolution was reduced to 0.5 m/px.”

Page 5 – Paragraph 3: Is the DEM from the UAV? Specify. Which are the implications of the error that the DEM has due to the flights (e.g. altitude of the flight, wind, reference points). Explain more

We derived the topographic correction for GPR cross-sections from the digital elevation model.”

Page 5 – Paragraph 3: Is there an image where you can show this? Perhaps you could combine the UAV flights and this?

I can see these are in image 4. Please add a reference to figure 4. That way readers can see what you mean

Five GPR profiles were taken in total, three of which…

Page 6 – paragraph 1: Did you test them all? Why? Justify and specify

“Testing of the algorithms and selection of the optimal ones can improve the output of palsa mires spatial variability and dynamics estim”

Page 6 – paragraph 3: Is this classification (TPI) the one you used? How is the variability of the slope in these terrains? Could you get the slope's differences well represented? Would be good to specify these things

Page 6 – paragraph 4: Is there any reason why did you use all LCC algorithms?

Page 6 – paragraph 4: Once NB defined, there is no need to define them again, just use initials

Page 6 – paragraph 4: References at the end of this paragraph

Page 6 – paragraph 5: How are these training sample collected? Are these ground truth? Is this visual interpretation? Need to be more specific

Page 6 – paragraph 5: Are ground control points independent from the training area. Please specify and explain clearly how you collected them, which are visual interpretations and which are field data

Page 6 – paragraph 5: Kappa coefficient is not a good estimator anymore. In the discussion you can add about the differences between the results obtained with the three classifiers and which are the implications of which one you use

Page 8, paragraph 1: is it a monotonic elevation trend? Why? Can you explain?

Page 8, paragraph 1: Would be interesting if you can explain how is the northern area in contrast with the south

were equivalent to most of the peat deposit area, except for the northern and eastern margins”

Figure 5: Add units to a and b panels; add what’s every initial of the veg types; there is a typo in the caption. It would be important to specify in the caption that these data is from the training sample. I would really like to see the same figure for the whole image (all pixels)

Page 10 – paragraph 1: Explain more. Why is that the reason that these are enough?

differences in morphometric variables between the classes were sufficient for automatic classification.”

Page 10 – paragraph 2: Were shadows a problem? Explain

Page 10 – paragraph 3: Why do you talk about the number of pixels resulted from SVM and not NB that was the best classifier? I think you need to be clear which results from the three classifiers are you using and be consistent over the text. If you will mix the classifiers, I would expect a final mosaic classification results including the best classes of the three classifiers. All these need to be explained in detail and clear. As it is now, it is a bit confusing.

Table 3: Kappa is not used anymore, this is a poor estimator of a classification.

Page 11: LLC is a typo (I suppose is LCC). Which land cover did you use for these results? NB, RF or SVM?

Page 12: From which classifier?

Figures 7 and 8 presented the longitudinal and transversal GPR cross-sections of the study site with the respective TWI values and land cover classes”

Page 12 - paragraph 1: Is it possible to see GPR signal characteristics plot by each LC?

Page 14 – paragraph 1: What about InSAR techniques? Some papers had shown the up and down movement and its correlation with vegetation classes. It is true that more research is needed and this study is quite important to show the relevance of using other types of data. But the take-away message, I think it is not that RS is not suffice. I think the take-away message can be much stronger and how ground penetrating radar is such a good tool to improve the detection of thawing but also could enormously help for understanding other techniques of EO remote sensing

Page 14 – paragraph 1: What about all inaccessible areas?

and subarctic and arctic territories can be a promising test ground, as they lack tall vegetation and can be surveyed by flying at low altitudes”

Page 14 – paragraph 2: I think this discussion should be more comprehensive as this is highlighted in your conclusions as one of your main findings

Page 15 – paragraph 1: Typo SfM, I suppose is SVM.

Page 15 – paragraph 2 – line 6: I agree. The permafrost inside palsas is variable, but in those lowest areas, where the palsas collapsed, permafrost is not occurring. Therefore, I would think that edges are still a good proxy for thawing monitoring over time. It can be quite powerful the combination of all these techniques to understand it better.

Page 15 – conclusions: It reads as if this sentence was the most important highlight in your paper. Do you think it is? Are these morphometric predictors for land cover or for permafrost thawing? Please specify

Our results highlight the morphometric predictors importance for machine learning since they improve the ac-curacy of LCC algorithms”

Page 15 – conclusions: This could be discussed further in the discussion

Our GPR data not only identified a high correlation with TWI values and LCC but also pinpointed some features of the vegetation cover shaped by the interior structure of the peat deposit and by permafrost position”

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

GENERAL COMMENTS:

I would like to congratulate the authors on presenting an interesting and novel approach towards monitoring palsa degradation. I believe that this work is significant for bridging the gap between ground-based methodologies, remote sensing and machine learning approaches; results from this paper highlight the importance of ground surveys for explaining spatial patterns of palsa thawing, where remote sensing technologies are unable to accurately account for the asymmetry of lateral erosion.

Overall I believe the paper is structured well, and I commend the authors for doing a very good job of documenting multiple varied analyses in a clear, sequential and interesting way. The paper is well-written and the figures are clear and impactful. The presentation of results is very good as it was easy to derive meaning from them. Overall, the methodology is simple to follow and adequately justified. A wide variety of materials and methods are included in this study to good effect to produce meaningful conclusions.

More detail documenting the geo-botanical surveys which inform the land cover classifications is needed. Firstly, more detail is required on the number of surveys per class, in particular. I appreciate that the number of surveys can be derived from the Error Matrix in the Supplementary Materials, however I think it is important to detail this information in the main text so that the reader can more easily interpret the quality of the train/test data and the model outputs themselves. I think it is also very important that the releve surveying approach is justified.

I support the publication of this paper in Remote Sensing with minor revisions that I have suggested in both the general and specific comments. Hopefully these suggestions will help to improve the clarity of aspects of the paper. I hope the authors will find them helpful.

SPECIFIC COMMENTS:

Study site: More justification of the releve sampling approach is needed here, as it strongly dictates the meaning of outputs of the land cover classification. Furthermore, more detail on the number of sites sampled per class would be good? This would also help the reader to better assess the quality of the output of the land cover classifications.

Figure 1: specify the units for elevation in the scale.

Figure 1 caption: replace ‘it’ with ‘its’.

Study site section: ‘The vegetation on hummocks and in flarks differs both as regards the key taxa composition and in the nutrition and moisture status’ – replace ‘as regards’ to ‘with regard to’.

Table 2: it would be helpful to know what your abbreviated land cover class names are referring to here. I see that these have been defined in the 5th paragraph of Section 3.3, but it helps to have the names within the table too – it helps provide context for readers who are not familiar with the geobotanical characteristics described in the table.

First sentence in results: ‘During the first stage, we analysed the imagery got by UAS photogrammetry’ – replace ‘got’ with ‘obtained’.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper is devoted to evaluating the correspondence between the surface traits of palsas degradation expressed in the landscape features and vegetation composition and the permafrost interior condition and peat hydrology based on remote sensing methods. From my personal standpoint, the paper is a very fine study that illustrate a possibility of successful using of various geoscience methods. Also, I guess you have showed, that the used methods in complex could become the methodological basis for monitoring programs of Arctic palsa ecosystems. 

Yet, I would like to recommend specify a tools or plugins that you have used for a land cover classification. The mere mention of QGIS and SAGA is not enough in my view. Firstly, it is interesting information for remote sensing specialists. Secondly, it will enable to estimate adequacy of methodology. Finally, it will help other scientists to repeat this methodology in their studies.

In my opinion that this paper could be published after adding this information.

 

Sincerely yours,

Reviewer

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Back to TopTop