Applicability of Relatively Low-Cost Multispectral Uncrewed Aerial Systems for Surface Characterization of the Cryosphere

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This article explores and confirms the effectiveness of UAVs in cryosphere research. It should be emphasized that fieldwork at high altitude is not easy and the obtained data are very precious. Although the language of this article is good and the pictures are very nice, I personally think that there is still a lot of potential for improvement and this article is not yet at the level of publication. I would like to suggest the following main questions about this study:

1. RTK technique was applied in this study. Is it possible for this study to explain further details of the RTK technique. For example, was a base station used and what kind of base station used?

2. The resolution of the individual UAV survey is presented in this study, but the positional accuracy of the different surveys should be provided. This means whether the ability to capture glacier change information is good or not.

3. Based on the accuracy, you should go ahead and utilize this valuable data and your excellent equipment to explore the glacier changes. Comparatively speaking, only demonstrating the potential for the application of UAV technology is a bit simple, as there are already enough studies that have demonstrated the potential of this technology for glacier and contribute to the glaciology community. For example, to investigate the movement and thinning patterns of glaciers, to characterize the development of glacial crevasses, and so on.

Author Response

Comment 1: RTK technique was applied in this study. Is it possible for this study to explain further details of the RTK technique. For example, was a base station used and what kind of base station used?

Response 1: The text in lines 224-241 detail the RTK technique that was used, how the base station was set up, the specific GNSS equipment that was used, and the number of ground control points that were used in each of the surveys to augment the positional accuracy achieved by the RTK surveys.

 

Comment 2: The resolution of the individual UAV survey is presented in this study, but the positional accuracy of the different surveys should be provided. This means whether the ability to capture glacier change information is good or not.

Response 2: The authors have added the Agisoft Metashape Structure from Motion processing reports for each of the UAV surveys to a new supplemental materials document which they have attached to the submission. This document contains details regarding the resolution and positional accuracy of the DEMs and orthomosaics and what parameters were used when constructing these products.  

 

Comment 3: Based on the accuracy, you should go ahead and utilize this valuable data and your excellent equipment to explore the glacier changes. Comparatively speaking, only demonstrating the potential for the application of UAV technology is a bit simple, as there are already enough studies that have demonstrated the potential of this technology for glacier and contribute to the glaciology community. For example, to investigate the movement and thinning patterns of glaciers, to characterize the development of glacial crevasses, and so on.

Response 3: The authors agree with your assessment that it would be informative to investigate the movement and thinning patterns of glaciers, to characterize the development of glacial crevasses, and so on. However, the scope of this study was to demonstrate the utility of a low-cost UAV multispectral system to a wide variety of cryospheric applications, as well as the glaciological community.  Future studies will further expand on the scientific applications. Our hope is that this study will be informative to other researchers who may be considering purchasing a UAV for cryospheric research applications. Furthermore, the imagery that we acquired of the Easton glacier is unsuitable for measuring ablation and thinning patterns since we were unable to place ground control points across the glacier surface due to the presence of crevasses and other hazards. Ground control points would have allowed us to achieve precise positional measurements across the glacier surface, which would have enabled us to measure thinning. The drone imagery by itself, however, is not precise enough to track these types of changes (there is too much error introduced when aligning images and generating the point cloud).

Reviewer 2 Report

Comments and Suggestions for Authors

 

The article presents three applications of the low-cost UAV, specifically the DJI Mavic 3 Multispectral on cryospheric studies at three US sites and compared it against pricier models like the DJI Matrice 350 at one of the sites. The article uses clear language and provides a detailed explanation and visualizations of how UAV techniques are employed in mapping snow algae, monitoring snowmelt and suncup formation, and classifying glacier surface features. It offers valuable insights into the use of affordable UAV technology in glaciology and snow science.

The authors have done a commendable job in providing the background, the current state of the field, and the goals of their research. The manuscript also stands out for its diversity in research design. The quality of the writing and presentation is high. I believe this research will interest other scholars and help others apply similar techniques to improve our understanding of the cryosphere. I recommend a minor revision. Below are my major comments, and I have attached a PDF with annotated detailed comments. Once these are addressed convincingly, I believe the manuscript could be accepted by the journal.

• In the Introduction, could you elaborate on how your research differs from other studies that also applied low-cost UAVs in glacier studies, such as Groos et al. 2019, Ewertowski et al. 2019, and others?
• A significant issue I noticed is that some objectives mentioned in the Introduction were not addressed. For example, snow depth (Line 77) and the retreat of the snowline (Line 80) were not discussed in the analysis or results. Lines 74-81 need careful review to ensure that the goals align with the subsequent analysis and case study results (please see detailed comments in the document).
• In Section 2.3, could you briefly mention the ownership of the study sites and whether/how permits were obtained for drone surveys in these areas?
• There is no validation of the red algae classification, especially since the spectral index method and the supervised random forest method produced very different results on total algae areas.
• A few suggestions for the figures: Please add scales to all panels in Figure 1, as well as a north arrow. I’d also suggest adding latitude and longitude marks on Figure 1a. Additionally, it would be helpful to mark the locations of GCPs on the Lemon Creek Glacier and Bagley Basin site figures where appropriate.

Comments for author File: Comments.pdf

Author Response

Comment 1: In the Introduction, could you elaborate on how your research differs from other studies that also applied low-cost UAVs in glacier studies, such as Groos et al. 2019, Ewertowski et al. 2019, and others?

Response 1: The authors added text to the introduction (lines 66-70) to clarify that the purpose of this paper is to assess the strengths and weaknesses of the Mavic 3M for glaciological research and to help inform other researchers who are interested in acquiring a UAV for their research decide if this particular model is suitable for their study questions or if they would benefit from a higher-cost system. Thus, the purpose of this research is more designed to be a broad overview of how low cost multispectral UASs can be applied to cryospheric research.

 

Comment 2: A significant issue I noticed is that some objectives mentioned in the Introduction were not addressed. For example, snow depth (Line 77) and the retreat of the snowline (Line 80) were not discussed in the analysis or results. Lines 74-81 need careful review to ensure that the goals align with the subsequent analysis and case study results (please see detailed comments in the document).

Response 2: The authors revised the text in the introduction (lines 77 to 83) to more accurately reflect the analysis and case study results that were presented in this article.

 

Comment 3: In Section 2.3, could you briefly mention the ownership of the study sites and whether/how permits were obtained for drone surveys in these areas?

Response 3: The authors added text to section 2.3 (lines 212-222) regarding ownership of the study sites and that no special permits were needed to fly in these locations.

 

Comment 4: There is no validation of the red algae classification, especially since the spectral index method and the supervised random forest method produced very different results on total algae areas.

Response 4: 30 snow algae samples were collected for another study as part of this field campaign. The authors used these sample locations to conduct an accuracy assessment since they serve as field verification for algal presence. The overall accuracy of each classification was calculated by dividing the number of algae samples that intersected with each of the classifications by the total number of algae samples. This methodology was included in Section 2.5 (lines 329 to 334), and the results were presented in Section 3.1 (lines 417 to 422) and in Table 1. Figure 5 and the ORG and supervised classifications were completely redone so that the area classified coincided with the area where algae samples were extracted from. The classified algal extent in Table 1 has therefore been updated as well.

 

Comment 5: A few suggestions for the figures: Please add scales to all panels in Figure 1, as well as a north arrow. I’d also suggest adding latitude and longitude marks on Figure 1a. Additionally, it would be helpful to mark the locations of GCPs on the Lemon Creek Glacier and Bagley Basin site figures where appropriate.

Response 5: The authors added scales and north arrows to both the study area figures. Figure 1 was converted into two separate figures to make it easier to understand. The Pacific Northwest is no longer included the reference panel. The locations of the GCPs have been added to the Lemon Creek figure. However, the locations of the GCPs at Bagley Basin changed during each survey, therefore they are not included in the figure.

Reviewer 3 Report

Comments and Suggestions for Authors

The article deals with a topic related to modern times. The article uses modern technologies, namely unmanned vehicles that capture multispectral data, which are able to detect phenomena invisible to the human eye. A large volume of obtained data was processed in the article, which indicates a great deal of work.

Subsequently, I recommend focusing on the following points:

The description of the flight parameters should be improved - the number of waypoints in individual planned flights and the flight time to cover the entire territory are missing.

The time of day for taking pictures using the UAV is not given.

Figure 1 is confusing; I recommend dividing it into 2 parts - the first with The Lemon Greek and the second with Mt. Baker and Bagley Basins

I recommend stating the spatial resolution of the resulting mosaic in the text (paragraph 2.4)

I do not consider it sufficient to use only 200 accuracy assessment points in such a large area for verification.

At first glance, Figure 3 seems like incomparable results, if possible, I recommend swapping the captured images

Describe the rationale for choosing different ORC values ​​for M3M and for M350micasense

In Table 1, do not indicate supervised classification, but directly indicate the selected random trees method.

Figure 7 is incomparable in the view of the area; it is necessary to unite the input images to the same area and only from these images to assemble mosaics and compare them

The drone parameters that are not described in the Methods and Materials section are listed in the discussion.

Author Response

Comment 1: The description of the flight parameters should be improved - the number of waypoints in individual planned flights and the flight time to cover the entire territory are missing.

Response 1: All flight surveys were planned within the DJI Pilot Pro 2 application, where the survey extent was specified by drawing a polygon on the map, and the flight lines and waypoints were automatically calculated. As such, the authors do not have metrics regarding the total number of waypoints. The authors added text describing the average flight times to cover each territory in Section 2.3 (lines 246 to 254).

 

Comment 2: The time of day for taking pictures using the UAV is not given.

Response 2: The time of day for taking pictures was added to Section 2.3 (lines 246 to 254).

 

Comment 3: Figure 1 is confusing; I recommend dividing it into 2 parts - the first with The Lemon Greek and the second with Mt. Baker and Bagley Basins

Response 3: The authors have divided the initial study area figure into 2 parts (Figure 1: Lemon Creek study area, Figure 2: Mt. Baker study area.

 

Comment 4: I recommend stating the spatial resolution of the resulting mosaic in the text (paragraph 2.4)

Response 4: The authors have listed the spatial resolution of each mosaic generated in this survey in Table A1 in Appendix A and have referenced this in Section 2.4 (lines 282 to 285).

 

Comment 5: I do not consider it sufficient to use only 200 accuracy assessment points in such a large area for verification.

Response 5: The authors have gone through and created 800 accuracy assessment points (200 per class), which is an increase of 4x over the initial number of accuracy assessment points. The overall accuracy did not change much based on the greater number of points (changed from 68% to 69%).

 

Comment 6: At first glance, Figure 3 seems like incomparable results, if possible, I recommend swapping the captured images

Response 6: The captured images were swapped.

 

Comment 7: Describe the rationale for choosing different ORC values ​​for M3M and for M350micasense

Response 7: The rationale for choosing different ORC values was added to Section 2.5 (lines 313-328).

 

Comment 8: In Table 1, do not indicate supervised classification, but directly indicate the selected random trees method.

Response 8: The classification method was changed to random trees.

 

Comment 9: Figure 7 is incomparable in the view of the area; it is necessary to unite the input images to the same area and only from these images to assemble mosaics and compare them

Response 9: The goal of this figure was not to show changes to a specific area over time, but to show the full area mapped with each survey. The text mentioned that a progressively larger area was mapped with each consecutive flight as we acquired additional batteries and methods were refined.

 

Comment 10: The drone parameters that are not described in the Methods and Materials section are listed in the discussion.

Response 10: The authors have revised the Methods and Materials section to include more of the drone parameters mentioned in other places in this paper.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Just from my perspective and the questions I asked earlier, I think the answers are basically acceptable. If the answers to the other reviewers are also considered reasonable, the editor may consider accepting it.