General Comment:

The paper by Ruols et al. presents an innovative application of drone-based ground-penetrating radar (GPR) technology to monitor the evolution of a near-terminus glacier collapse feature on the Rhône Glacier, Switzerland. The authors study a high-spatial-density 4D GPR dataset collected during four surveys in summer-autumn 2022. It focuses on the formation and temporal evolution of a subglacial air cavity and associated drainage systems. The authors demonstrate how the interplay of subglacial hydrology, mechanical failure, and inflow of warm air led to the development of circular surface crevasses and eventual roof collapse.

Building on methods developed in earlier work by a similar author team, this study achieves a full 4D reconstruction of a near-terminus glacier collapse feature. Considering the rarity of 4D GPR studies in the literature, particularly using drones over glaciers, this work is undoubtedly valuable and within the scope of The Cryosphere. Moreover, the focus on glacier snout collapse—a process increasing due to global warming—places the study in a timely and broader context.

By leveraging drone-based GPR, the authors overcome limitations of traditional GPR acquisition in rough terrain, with advantages for dataset resolution, acquisition speed, and safety. Drone-based ground-penetrating radar (GPR) technology has emerged as a transformative tool for subsurface exploration across various fields, including glaciology. Importantly, the results have implications for understanding glacier instabilities and their responses to external drivers, such as climate change.

Overall, the study is well-structured and clearly written, with figures and supplementary material that effectively support the narrative. However, there are areas where minor improvements could enhance the manuscript, particularly in expanding the discussion of uncertainties, and further contextualizing the study within a broader glaciological-climatological framework. I hope that my comments help the authors improve the manuscript. 

Specific comments:

Abstract

I suggest slightly reducing the detailed results in the abstract and focusing on a concise summary of the findings. This would better emphasize the innovation of the applied methodology as a key tool, enabling new insights into glaciological processes, strengthening the connection between the method and the results.

Introduction

The introduction effectively sets the stage, but a broader explanation of the advantages and limits of 4D drone/helicopter-based GPR compared to other ground-based methods would strengthen the impact. For example, discussing how non-ground-based systems improve safety, enable data collection in inaccessible or hazardous areas, and allow for high-resolution, repeated surveys over short timescales could highlight their importance for advancing glaciological research. At the same time, acknowledging potential limitations, such as challenges in vertical positioning accuracy due to glacier surface changes or lateral obstacles as well as the interaction between the 3D GPR signal lobes with the topography (e.g. see Forte et al., 2019), would provide a balanced perspective and enhance the methodological transparency of the paper.

l51: Consider adding “and in proximity of strong lateral reflectors (Forte et al. 2019).”

• Forte, E., Basso Bondini, M., Bortoletto, A. et al.Pros and Cons in Helicopter-Borne GPR Data Acquisition on Rugged Mountainous Areas: Critical Analysis and Practical Guidelines. Pure Appl. Geophys. 176, 4533–4554 (2019). https://doi.org/10.1007/s00024-019-02196-2

ll54-61: There is a slight imbalance in the details provided for Ruols et al. (2023) compared to the other works (Jenssen et al., 2020; Tan et al., 2021; Valence et al., 2022). Rephrasing this section and including recent works such as Tjoelker et al. (2024) and SelbesoĞlu et al. (2023) would create a more balanced discussion.

• Tjoelker, A. R., Baraër, M., Valence, E., Charonnat, B., Masse-Dufresne, J., Mark, B. G., & McKenzie, J. M. (2024). Drone-Based Ground-Penetrating Radar with Manual Transects for Improved Field Surveys of Buried Ice. Remote Sensing, 16(13), 2461. https://doi.org/10.3390/rs16132461
• SelbesoĞlu, M. O., Karabulut, M. F., Oktar, Ö., Akpinar, B., Vassilev, O., Arkali, M., ... & Özsoy, B. (2023). Accuracy assessment of glacier depth monitoring based on UAV-GPR on Horseshoe Island, Antarctica.Turkish Journal of Earth Sciences, 32(8), 999-1012.

l92 Please, be more specific about “recently” as lakes formed during 1990s and in 2005 according to Tsutaki et al. (2013).

I also suggest providing some more context concerning glacier changes (including collapse) in relation to the ongoing climate change.  

Methods (acquisition and processing)

While the authors often refer to a previous paper (Ruols et al. 2023) for methodological details, adding a few more information directly in this manuscript would improve clarity and accessibility. Unless I missed it, I suggest adding some information about:

• the drone set-up
  • l112: is it a shielded antenna?
  • A zoomed picture of the drone-GPR system could be added as a box to Fig. 3
• the development of the flight plan
  • I did not understand if/how cross-profile were acquired or interpolated, as in Fig. 2 the acquisition seems to be done along parallel profiles only, but in Fig.6 both inline and crossline profile are shown).
• How the 3D GPR signal lobes interact with the topography in particular in relation with the height above the ground and the angle between the GPR and the surface?

Fig 2 and Table 1 - I might be wrong, but I think photogrammetry was never mentioned before (or after) in the text. Even if the acquisition of the orthophotos was carried out by ETH Zürich’s VAW Glaciology group, you should mention in the main text how you use this dataset and provide some details about it.

Please, when possible, substitute general statements with quantitative ones:

• L118: “with a high level of repeatability for the horizontal positioning (Fig. 2e).” Can you quantify this?
• l119: “However, differences in vertical positioning between acquisitions were present due to glacier melting” Do you mean that the altitude above sea level has changed, but the altitude above the surface is always 5 m since it is controlled by the True Terrain Following? Can you quantify this change?
• Ll120-122 – “Advantages of a drone- based GPR acquisition are clear, as high-density data could not have been acquired on the glacier surface because of the large crevasses.“ What are these advantages? E.g., safety, time, difficulty in pulling a ground-based GPR over rough terrain…

Ll191-192: Considering the highly heterogeneous case study with ice, air and water, what is the associated error of using a single velocity? Can it be estimated, perhaps using bedrock depth from the four acquisitions?

Results and Discussions

ll274-275: The picking process for the air cavity should be introduced earlier in the methods.

Table 2: Would it be possible to provide an error for the measurements?

ll302-304 - “Regarding the two subglacial channels, the main one, originating from the northeast, is likely to drain the majority of the glacier’s subglacial water system, whereas the second one, originating from the southeast, likely drains a constrained hydrological basin on the orographic left-hand side of the glacier.“ Could you provide some information to explain why you think this?

Broader implications

While the paper provides a thorough examination of a specific glacier collapse, it could enhance its impact by more explicitly contextualizing this phenomenon within the broader framework of global warming. Currently, the connection between the findings and global warming is only briefly mentioned through the reference (Egli et al., 2021b). While it is clear that a single collapse event cannot be directly attributed to the ongoing climate change, the increasing frequency of such events is linked to rising temperatures. Adding one or two sentences to address this point would help draw attention to the broader relevance of glacier snout collapses, which are not only indicative of cryospheric changes but can also have significant implications for human safety in mountain environments. This discussion could be incorporated into the Discussion or Conclusions sections, highlighting the importance of monitoring these phenomena in the context of climate-driven hazards.

Technical comments:

ll62-64: I suggest moving this paragraph after the discussion on terminal collapses (L75) to consolidate all relevant content in one section.

L95: The reference to "boxes b-c" and "d-e" in Figure 1 could be clarified by separating these into distinct references for each sentence.

Fig1: Consider making box (a) as wide as boxes (b+c) and highlighting the crevasses and collapse features in boxes (b), (c), and (d).

Ll146-149: The first 3 sentences fit more into the acquisition section. I suggest moving them.

l82 changes in  changes of

L189: Wasn’t the height 5 m above the surface?

L93 tongue --> terminus

L231 sentences --> paragraph

L238 Please define DOP (it was defined in the label of Fig. 2, but should be defined also in the main text).

Ll262-263: This sentence fits more in the methods section than in the results.

Fig 9.: “Elevation” in the y-axis label could be repeated only once per side.

To my knowledge this study is the first one to date that performs detailed 4D GPR measurements for a glacier collapse feature.

Such collapse features have been occurring with increasing frequency on glaciers in the Alps since the early 2000s in the context of accelerated glacier retreat, but also elsewhere on the planet. Therefore, this study makes an important contribution to the study of rapid glacier changes in a warming climate, attempting to fill a knowledge gap concerning englacial and subglacial mass loss, but also testing a methodology that his rather new and that has great potential in glaciology – especially for the study of sub- and englacial hydrological features, but also for tracking internal reflectors and other 3D structures. While the application of high-precision 3D GPR on glaciers has been tested previously in several studies, 4D studies are rather new, especially when making use of drones and RTK positioning.

General comments

Both the field data acquisition and data analysis seem to have been executed with great care and are well documented. The precision and detail of the data acquisition are remarkable and should set a high standard for future studies using drone-based GPR with RTK positioning.

The results are presented in an accessible manner, and they nicely document the rapid changes occurring in the growing cavity underneath Rhône glacier as well as its outlet stream. GPR data was combined with orthophotos, Digital Elevation Models and borehole camera observations as well as photographs and visual observations.

The methods used for processing GPR data are based on existing literature and are easy to follow by the instructed reader.

One area where a more in-depth discussion could be added is on the potential role of sediment and sediment erosion in the initiation and widening of the subglacial cavity. It could be discussed if the glacier bed underneath the cavity is only composed of bedrock, or if there is or was a sediment layer that has been eroded or accumulated during summer 2022.

Further, a few details about the drone GPR survey could be added, such as the recording frequency of the GPR. But it is clear that several details are already explained in a previous publication by the same author, focusing on the drone radar system and data acquisition.

I recommend this manuscript for publication after minor revisions.

Detailed comments

L 83 : Is this (2022) the latest GLAMOS reference for Rhone?

L 92: word order : “…investigated by Church et al. (…) using GPR to…”

Figure 1, L 99: which type of satellite image / source?

L 111: Impressive antenna. What is its weight? (“featherweight”)

L 112: “transmitter-receiver”?

L 116: Did you conduct tests for along-glacier-flow direction? (asking out of curiosity)

L 119: Mostly out of curiosity: Would it have been useful to try and maintain a similar flying height as the previous flight despite glacier melting (e.g., changing the height above the ice, or using the flying heights of previous flights)? Or would that change the signal too much as the distance between antenna and ice increases, and the coupling to the ice surface therefore changes?

Figure 2: (e) impressive positioning precision between different dates. (f) Maybe name the y-axis “acquisition elevation” for clarity?

L 144: Maybe elaborate a bit more, in 1-2 additional sentences?

L 151: Was the same recording frequency used as in Ruols et al. (2023)?

L 152: You might want to explain that several flights were needed to change batteries. Knowing that this is explained in Ruols et al. (2023) as well.

L 160: What was the GPR recording frequency?

L 190: You might want to provide 2-3 specific original references justifying the chosen velocity of 0.167 m ns-1.

Figure 4 / L207: “..325 ns (purple)” : I see this as blue.

Figure 5: It might help to add a legend for the blue, yellow, red arrows in the figure.

Figure 6: What features or situation can we see in the depth slices g, h, i?

L 221: Remove “Indeed, ”

L 234: Could this consideration of maximum reflection strength over a 2-m window introduce some sort of bias or artefact?

L 238: Did you ever introduce “DOP” (Digital Orthophoto) ?

Figure 8: Maybe a detail, but still worth mentioning for future / further investigation: There is a strong high amplitude signal visible in the lower corner of each plot (25N / 130 E), maybe indicating the edge of another channel, or ponding. Alternatively, it could be an artefact, as it is on the edge of the dataset.

Also figure 8, L 252: “..leaving the feature westwards..”

L 273: “….due to a combination of ice creep into the cavity and partial mechanical failure.” Maybe be a bit more careful with this statement and present it as a hypothesis?

L 310: “..over time”

L 321: “…evolves throughout summer “

L 328: Mechanical failure (and erosion of subglacial till) was, among others, also hypothesized by Egli et al. (2021b), but under the name of “block caving” (Paige, R. (1956). Subglacial stoping or block caving: A type of glacier ablation. Journal of Glaciology, 2, 727–729. https://doi.org/10.3189/s0022143000024977). Very similarly to Rhône, ice blocks floating out of the terminus at Otemma were observed already in summer 2017 – the year before the ice surface collapse event. But no borehole was made to verify if a cavity had started to form while the glacier outlet channel was still pressurized. The correct main finding for Rhône remains that the outlet channel at Rhône seems to have remained pressurized for several weeks while a large cavity was forming underneath the ice.

L 333: This is an interesting and valuable discussion. You could talk a bit more about other potential mechanisms for channel widening and cavity opening, namely sediment erosion (and deposition). Did you determine whether the ground below the collapse feature is / was mainly composed of bedrock, or also sediments? Or, if there used to be sediments, but they were eroded away by the subglacial channel during the formation of the collapse feature?

This also raises the question about what initiated the formation of the first cavity, making flow non-pressurized, and which then led to roof destabilization, detachment of lamellae, etc., to start with?

Figure 11: There are lots of (partially eroded) sediments, and bedrock, visible in this picture. You should talk about this in the discussion, and about the sediments’ potential role in the initial formation of the cavity.