Subaerial and subglacial seismic characteristics  of the largest measured jökulhlaup from  the eastern Skaftá cauldron, Iceland

Eibl, Eva P. S.; Vogfjörd, Kristin S.; Ófeigsson, Benedikt G.; Roberts, Matthew J.; Bean, Christopher J.; Jones, Morgan T.; Bergsson, Bergur H.; Heimann, Sebastian; Dietrich, Thoralf

doi:10.5194/esurf-11-933-2023

Articles | Volume 11, issue 5

https://doi.org/10.5194/esurf-11-933-2023

© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/esurf-11-933-2023

© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 11, issue 5

Research article

|

05 Oct 2023

Research article |

| 05 Oct 2023

Subaerial and subglacial seismic characteristics of the largest measured jökulhlaup from the eastern Skaftá cauldron, Iceland

Eva P. S. Eibl, Kristin S. Vogfjörd, Benedikt G. Ófeigsson, Matthew J. Roberts, Christopher J. Bean, Morgan T. Jones, Bergur H. Bergsson, Sebastian Heimann, and Thoralf Dietrich

Download

Final revised paper (published on 05 Oct 2023)
Preprint (discussion started on 20 Jul 2022)

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2022-644', Anonymous Referee #1, 22 Aug 2022

Review to “Seismic Characteristics of the Largest Measured Subglacial Flood from the Eastern Skaftá cauldron, Iceland” by Eibl and others, 2022

The authors of the manuscript present seismological observations from a subglacial flood that originated in 2015 from the Eastern Skaftá cauldron at the Vatnajökull ice cap in Iceland. Over the course of a few days, they detect various quakes and two types of tremors which they exploit to study the temporal evolution of the flood by means of quake locations, beamforming analysis, spectrograms and tremor amplitude. Guided by complementary measurements of the ice motion and hydrological parameters, Eibl et al. find that the quakes are related to the subsidence of the cauldron and the two tremor types to the subglacial hydrology: the longer lasting type 1 tremor is attributed to the flood wave propagation while the shorter but more impulsive type 2 tremor is attributed to geothermal processes including boiling in response to the flood. The study focuses on type 1 tremor and the authors suggest that this signal is caused by multiple brittle ice cracking as the flood wave propagates between ice and bedrock, lifting the glacier by up to 1 m. This tremor type can be used to track the propagation of the flood and its speed via beamforming analysis (Eibl et al. 2020).

The investigated processes are interesting and highly relevant for the glaciological(/volcanological) community. Furthermore, the results appear robust and the conclusions convincing. However, in my opinion, the submitted manuscript does not provide significant new data and/or processing and thus not significant novel insights into the flood. Most of the eight figures show material, that is already presented in Eibl et al. (2020) (e.g. most of Figs. 1, 3, 4, 6) and the conclusion drawn are the same for both manuscripts. Doubtlessly, the present manuscript contains a more in-depth discussion on the involved processes compared to Eibl et al. (2020), but still concludes that type 1 tremor is caused by the propagating flood wave lifting the ice and type 2 tremor by hydrothermal explosions and subsequent geothermal boiling. Only the quake and water-chemical data are newly introduced but these solely play a side role and do not provide significant novel insights. Overall, the manuscript appears more like a supplementary material to Eibl et al. (2020). For this reason, I unfortunately cannot recommend to consider the article for publication in Earth Surface Dynamics.

Reference

Seismic ground vibrations give advanced early-warning of subglacial floods. 11, 2504 (2020). https://doi.org/10.1038/s41467-020-15744-5

Citation: https://doi.org/10.5194/egusphere-2022-644-RC1
- AC1:
  'Reply on RC1', Eva Eibl, 30 Nov 2022
  We thank the reviewers for their considered and insightful review. We agree with the reviewers that the processes are interesting and highly relevant for the glaciological and volcanological community when it comes to further our understanding of hazardous flooding events but also the heat exchange between rock and ice.
  Based on the feedback, we recognize that we did not draw sufficient attention to the new insights that this manuscript is focusing on in comparison to Eibl et al. 2020. To improve the manuscript and address the reviewer's concerns, we have improved clarity throughout in the old text and figures, to draw attention to the new findings and remove duplication.
  We have also conducted new processing of the seismic data from the SIL network in combination with the seismic arrays at JO and IE. We have now processed (i) all three components of the arrays. (ii) We performed template matching and clustering of additional icequakes that indicate the collapse of the ice cap above the lake. (iii) We detected high frequency transient events in the array data that support our interpretation that Type 1 tremor is composed of repeated events. (iv) We investigated the ‘local noise’ in more detail and located the sources at rapids and a nearby waterfall.
  In comparison to Eibl et al. 2020, we here show that we can differentiate 4 instead of 2 different tremor types. Our detailed seismological analysis yields 3 different subglacial tremor types (Type I to III), rapid-induced tremor (Type 4) from subaerial river flow, icequakes originating in the cauldron area, and high-frequency transient events propagating alongside Type I tremor with the flood front. This study hence provides much more detail of the detected seismological signals and their generation associated with the subglacial but also subaerial propagation of the flood.
  As recognized by the reviewers we would like to help to classify tremors in less studied areas with this detailed and extensive study of a robust dataset. Further details of our modifications can be found in bullet points below:
  We performed a new array processing at higher frequencies and detected transient events that follow the flood front. This supports our conclusion that tremor Type 1 is composed of icequakes.
  
  We ran a STA/LTA detector and correlated the waveforms to detect more events. Our final catalog was clustered based on the station HAM, which was the closest 3-component seismometer to the sources based on our back azimuth estimate. Based on station HAM, the waveforms were clustered into 20 different families. These all originate in the cauldron area indicating a gradual collapse of the ice-shelf on different faults.
  
  We analyzed, located, and discussed the other detected seismic sources which are caused by the flood once it reaches the subaerial river (previously just referred to as "local noise" in Eibl et al 2020)
  
  We realized that we had not made it clear enough that we are discussing three different tremor types here (while Eibl et al. 2020 only discuss two)
  
  We modified all figures to highlight the above-mentioned new points (and deleted former Fig. 2 as this was merely doubling most information).
  
  We clarified that the role of the geochemistry is crucial when it comes to discussing the source of Type 2 and 3 tremor in the context of volcanic eruptions or hydrothermal explosions.
  
  We modified the discussion to add our new insights about the drainage processes.
  
  Based on our additional analysis we modified our abstract and conclusion.
  
  Citation: https://doi.org/10.5194/egusphere-2022-644-AC1
RC2:
'Comment on egusphere-2022-644', Anonymous Referee #2, 02 Sep 2022

This preprint describes the largest measured subglacial flood from the Eastern Skafta cauldron in Iceland in 2015. The Authors aim to improve the current understanding of processes behind seismic signal generation during subglacial floods. Thanks to the analysis of seismic, GPS, and hydrological observations, the Authors propose two source mechanisms from tremor signal generation: geothermal boiling of water in crustal rocks and repeating icequakes caused be glacier lift.

Yet, most of these observations and the same event have been already published in the paper by Eibl et al., 2020. Moreover, the Authors used the same methods to analyze seismic data. I believe that for this paper to be published, more new information or novel processing approaches should be explored. Some of the claims seem speculative now; for example, the authors propose that tremor 1 is associated with repeating icequakes. This can be very easily verified with clustering methods (e.g., RedPy, Hotovec-Ellis et al., 2019) or template matching (Beaucé et al., 2018). For now, I do not see much value added and novelty compared to Eibl et al. 2020 paper, which, unfortunately, does not allow me to accept this preprint.

References:

Beaucé, E., Frank, W. B., and Romanenko, A.: Fast Matched Filter (FMF): An Efficient Seismic Matched-Filter Search for Both CPU and GPU Architectures, Seismological Research Letters, 89, 165–172, https://doi.org/10.1785/0220170181, 2018

Eibl, E.P.S., Bean, C.J., Einarsson, B. et al. Seismic ground vibrations give advanced early-warning of subglacial floods. Nat Commun 11, 2504 (2020). https://doi.org/10.1038/s41467-020-15744-5

Hotovec-Ellis, A. and Jeffries, C.: Near Real-time Detection, Clustering, and Analysis of Repeating Earthquakes: Application to Mount St. Helens and Redoubt Volcanoes, in: Presented at Seismological Society of America Annual Meeting, 2016

Citation: https://doi.org/10.5194/egusphere-2022-644-RC2
- AC2:
  'Reply on RC2', Eva Eibl, 30 Nov 2022
  We thank the reviewers for their considered and insightful review. We agree with the reviewers that the processes are interesting and highly relevant for the glaciological and volcanological community when it comes to further our understanding of hazardous flooding events but also the heat exchange between rock and ice.
  Based on the feedback, we recognize that we did not draw sufficient attention to the new insights that this manuscript is focusing on in comparison to Eibl et al. 2020. To improve the manuscript and address the reviewer's concerns, we have improved clarity throughout in the old text and figures, to draw attention to the new findings and remove duplication.
  We have also conducted new processing of the seismic data from the SIL network in combination with the seismic arrays at JO and IE. We have now processed (i) all three components of the arrays. (ii) We performed template matching and clustering of additional icequakes that indicate the collapse of the ice cap above the lake. (iii) We detected high frequency transient events in the array data that support our interpretation that Type 1 tremor is composed of repeated events. (iv) We investigated the ‘local noise’ in more detail and located the sources at rapids and a nearby waterfall.
  In comparison to Eibl et al. 2020, we here show that we can differentiate 4 instead of 2 different tremor types. Our detailed seismological analysis yields 3 different subglacial tremor types (Type I to III), rapid-induced tremor (Type 4) from subaerial river flow, icequakes originating in the cauldron area, and high-frequency transient events propagating alongside Type I tremor with the flood front. This study hence provides much more detail of the detected seismological signals and their generation associated with the subglacial but also subaerial propagation of the flood.
  As recognized by the reviewers we would like to help to classify tremors in less studied areas with this detailed and extensive study of a robust dataset. Further details of our modifications can be found in bullet points below:
  We performed a new array processing at higher frequencies and detected transient events that follow the flood front. This supports our conclusion that tremor Type 1 is composed of icequakes.
  
  We tested RedPy, but the clustering based on a predefined transient list did not yield any families. We hence ran a STA/LTA detector and correlated the waveforms ourselves to detect more events. Our final catalog was clustered based on the 3-component station HAM, which was closest to the sources based on our back azimuth estimate. Based on station HAM, the waveforms were clustered into 20 different families. These all originate in the cauldron area indicating a gradual collapse of the ice-shelf on different faults.
  
  We analyzed, located, and discussed the other detected seismic sources which are caused by the flood once it reaches the subaerial river (previously just referred to as "local noise" in Eibl et al 2020)
  
  We realized that we had not made it clear enough that we are discussing three different tremor types here (while Eibl et al. 2020 only discuss two)
  
  We modified all figures to highlight the above-mentioned new points (and deleted former Fig 2 as this was merely doubling most information).
  
  We modified the text severely to reflect the additional processing we have done.
  
  Citation: https://doi.org/10.5194/egusphere-2022-644-AC2
RC3:
'Comment on egusphere-2022-644', Anonymous Referee #3, 22 Sep 2022

The manuscript “Seismic Characteristics of the Largest Measured Subglacial Flood from the Eastern Skaftá cauldron, Iceland'' presents seismic, GPS and hydrological data, collected during a subglacial flood event in Iceland. The authors found two different types of glacial tremors. Both are related to the flood event. Type 1 tremor originates due to the glacier uplift, leading to icequakes. In the manuscript, the authors show the propagation of the tremor towards the glacier terminus by calculating the back azimuth and comparing these results with GPS and hydrological measurements. Additionally to the tremor caused by icequakes (type 1), a second tremor type caused by boiling water, called type 2, is presented. After emptying the subglacial lake, the pressure at the glacier bed decreases and allows the water in the volcanic hydrothermal system to boil. Seismometers can see the exploding water bubbles, which show up as type 2 tremor. The authors also present the typical frequency band of both tremor types.

The study of interactions between solid, fluid and gaseous water is unique. In my opinion, the topic of this manuscript is important to understand the hazardous flooding events but also the heat exchange between rock and ice better. Nevertheless, to me the manuscript does not include enough new findings, compared to Eibl et al., 2020, to be presented as a self-standing paper. The manuscript seems to me more like supplementary information to Eibl et al., 2020 because most of the figures and the results are already presented in Eibl et al., 2020. Based on my findings, I do not recommend this manuscript for publication in the current state.

Adding new methods to the same dataset would help to make this manuscript a self-standing paper. For example, Eibl et al., 2020 proposed an early-warning approach using tremor type 1, which could be tested in this manuscript. Increasing the warning time and reaching as few false alarms as possible would be a beneficial application of the findings shown. Another further application would be to train machine learning algorithms to cluster the tremor types automatically by analyzing different tremor features in the time and frequency domain. The seismic dataset seems robust and can help to classify tremors in less studied areas. Also trying to reproduce the measured seismic data with a physical model would improve the manuscript and make it a self-standing paper. A physical model would help to understand the complicated processes at the glacier ice, glacier bed, and englacial lake interface, including the volcanic heat fluxes. However, based on the above-mentioned reasons I can not recommend accepting this manuscript for publication in Earth Surface Dynamics.

Reference:

Eibl, E. P. S., Bean, C. J., Einarsson, B., Pàlsson, F., & Vogfjörd, K. S. Seismic ground vibrations give advanced early-warning of subglacial floods. Nature Communications 11, 2504 (2020). https://doi.org/10.1038/s41467-020-15744-5

Citation: https://doi.org/10.5194/egusphere-2022-644-RC3
- AC3:
  'Reply on RC3', Eva Eibl, 30 Nov 2022
  We thank the reviewers for their considered and insightful review. We agree with the reviewers that the processes are interesting and highly relevant for the glaciological and volcanological community when it comes to further our understanding of hazardous flooding events but also the heat exchange between rock and ice.
  Based on the feedback, we recognize that we did not draw sufficient attention to the new insights that this manuscript is focusing on in comparison to Eibl et al. 2020. To improve the manuscript and address the reviewer's concerns, we have improved clarity throughout in the old text and figures, to draw attention to the new findings and remove duplication.
  We have also conducted new processing of the seismic data from the SIL network in combination with the seismic arrays at JO and IE. We have now processed (i) all three components of the arrays. (ii) We performed template matching and clustering of additional icequakes that indicate the collapse of the ice cap above the lake. (iii) We detected high frequency transient events in the array data that support our interpretation that Type 1 tremor is composed of repeated events. (iv) We investigated the ‘local noise’ in more detail and located the sources at rapids and a nearby waterfall.
  In comparison to Eibl et al. 2020, we here show that we can differentiate 4 instead of 2 different tremor types. Our detailed seismological analysis yields 3 different subglacial tremor types (Type I to III), rapid-induced tremor (Type 4) from subaerial river flow, icequakes originating in the cauldron area, and high-frequency transient events propagating alongside Type I tremor with the flood front. This study hence provides much more detail of the detected seismological signals and their generation associated with the subglacial but also subaerial propagation of the flood.
  As recognized by the reviewers we would like to help to classify tremors in less studied areas with this detailed and extensive study of a robust dataset. Further details of our modifications can be found in bullet points below:
  We think that testing the early-warning approach would bring this manuscript closer to Eibl et al. 2020 and hence did not implement this.
  
  We also think that a sample size of 4 is too small to implement a systematic early warning approach or machine learning.
  
  However, we performed a new array processing at higher frequencies and detected transient events that follow the flood front. This supports our conclusion that tremor Type 1 is composed of icequakes.
  
  We ran a STA/LTA detector and correlated the waveforms to detect more events. Our final catalog was clustered based on the station HAM, which was the closest 3-component seismometer to the sources based on our back azimuth estimate. Based on station HAM, the waveforms were clustered into 20 different families. These all originate in the cauldron area indicating a gradual collapse of the ice-shelf on different faults.
  
  We analyzed, located, and discussed the other detected seismic sources which are caused by the flood once it reaches the subaerial river (previously just referred to as "local noise" in Eibl et al 2020)
  
  We realized that we had not made it clear enough that we are discussing three different tremor types here (while Eibl et al. 2020 only discuss two)
  
  We modified all figures to highlight the above-mentioned new points (and deleted former Fig 2 as this was merely doubling most information).
  
  We modified the text throughout to reflect these additions.
  
  Citation: https://doi.org/10.5194/egusphere-2022-644-AC3

Peer review completion

AR – Author's response | RR – Referee report | ED – Editor decision | EF – Editorial file upload

AR by Eva Eibl on behalf of the Authors (30 Nov 2022) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (26 Dec 2022) by Daniella Rempe

RR by Anonymous Referee #2 (10 Feb 2023)

Suggestions for revision or reasons for rejection

In the revised version of the manuscript, the Authors provide an additional analysis of the signals seismic signal generation during the largest measured subglacial flood from the Eastern Skafta cauldron in Iceland in 2015.

Now the manuscript provides some additional information compared to Eibl et al., 2020. However, the description of different processes associated with subglacial floods and the corresponding seismic signals is not clear. Also, the description of methods could be more precise, especially since the method section is mixed with the results. In general, the manuscript is interesting but not easy to follow, and I suggest that the Authors spend some time clarifying and simplifying specific parts of the manuscript.

I have a few major comments. First, it needs to be clarified how the Authors divide the seismic signals into different types of tremors. The Authors use different seismic metrics (e.g., back azimuth, slowness, frequency content) combined with ancillary measurements (e.g., GPS, hydrological) to classify the seismic signals into tremors 1, 2, 3, and 4. However, this classification seems somehow arbitrary. I believe it would be easier for the Reader to follow the manuscript if the Authors first introduced different metrics they use to describe the seismic signals and constrain their sources; based on those metrics and ancillary data, divide seismic signals into different types of “tremors,” and then finally interpret the tremors in terms of source processes generating them.

Then, the methods, results, and interpretation are all mixed. Certain parts of the manuscript could be shortened (e.g., 5.2.1), and a certain part gives repeating information (see below for details). Moreover, the Authors used multiple terms to describe the same methods (e.g., array processing, beamforming, FK…) and the same processes (e.g., quakes, icequakes, and transient events)-maybe their signals have different source processes, but for now, it is very hard to keep track of which signals are generated by what.

Then, some terms do not seem fully accurate. For example, the Authors call tremor 3 a harmonic tremor. The spectrogram in Figure 7b does not show any spectral lines that I would expect to see in a harmonic tremor. Also, the amplitude peak in PSD in panel g for the dominant frequency does not seem very pronounced compared to panel e (non-harmonic termor, I assume?). Also, what is the cross-correlation coefficient between “transient events”? 0.2? If yes, you cannot really say these are repeating events or event multiplets… See Uchida and Bürgmann (2019).

See below for more comments:

Line 40: When studying… - Why do those different sources need to be characterized? Could you add some more information on the impact of this characterization?

Line 65: Could you give the exact days?

Line 114: Is it possible that those 45 events are the events that you also detect through STA/LTA and template matching?
Line 128: If you use the horizontal components, the FK should be performed on rotated components, R and T; otherwise, the slowness measurements are affected. Also, do you only show the results for the vertical component in the manuscript, or did I miss something?

Line 131: At less… -this should be in results
Line 133: Define semblance.

Line 139: This belongs in Results.

Line 147: That would fit better in Results again.
Line 152: FK is also an array processing method… In fact, what you describe here is also referred in the literature as FK (the search over sx and sx, if I understand correctly). Could you show some results of that in an appendix?

Line 270: on the lower half of the flood propagation path-not clear
Line 180: The threshold of 0.2 seems too low to claim a similarity of waveforms.
Line 189: the closest to the flood path?

Line 199: We selected…-this sentence is very confusing. You did not mention beam power before.

Line 214:Result and Interpretation-isn’t Interpretation Discussion?

Line 232: a significant amount-could you specify?
Line 239: conductivity measurements were not introduced yet?

Line 246: Quakes or icequakes?
Line 291: You already said it.
Line 295: Again, are those events really “repeating” and highly similar? What is the cross-correlation ratio between these events?

Line 299: similar characteristics-such as?
Line 305: in the range of 0.8 to 0.8 s/km

Lines 387:392-this has already been said in Introduction.

Figure 1: Gray vertical line at 20h; what does it denote? Can some of those quakes also be classified as transients? Could you show some waveforms?
The black curve shows forward-calculated(?) changes…

Figure 3: You do not mention beamforming in the text.

Figure 5: Could you annotate the vertical gray lines? Any comments on the change between slowness values of different clusters? The one appearing earlier have low slowness values ~0.2 s/km; could they be coming from depth? Also, panel c shows a progressive activation of different clusters. That seems interesting. Could you comment on this?

Figure 7: As I already said, is this really a harmonic tremor? Could you try to improve the quality of the spectrogram in b to check if any spectral lines are visible? Panels c, d, e, f, and g are missing labels and units (c, e, and g).

Figure 8: Again, the amplitude peaks at dominant frequencies are not visible for the “harmonic” termor.

Uchida, N., & Bürgmann, R. (2019). Repeating earthquakes. Annual Review of Earth and Planetary Sciences, 47 (1), 305-332. Retrieved from https://doi.org/10.1146/annurev-earth-053018-060119 doi: 10.1146/annurev-earth-053018-060119

Hide

RR by Anonymous Referee #4 (15 Jun 2023)

Suggestions for revision or reasons for rejection

In fairness to the authors and in keeping with prior discourse on this manuscript, I have tried to focus my review on how well the authors have addressed the issues that all three previous reviewers identified. In broad strokes, I would agree with their assessment. The dataset is interesting and offers some intriguing possibilities toward understanding outburst flood processes, hydrothermal/ice interactions, and hazard early warning. I am not a seismologist and am approaching this from a geomorphologist’s perspective so I can’t offer too much on methodology, and the previous reviewers seem to agree that it’s all good.

I see some substantive improvement particularly in the discussion of tremor sources and origins. I don’t know that I would say that splitting the type-2 tremors into 2 and 3 referring to the tremor generated by hydrothermal explosions and subsequent boiling is new, as the type-2 tremor is describes in Eibl et al. (2020) as likely being driven by both processes already, but in general, I think the additions make the manuscript a lot more distinctive from the 2020 paper. I see a few issues that I think the authors could address without much new analysis.

- First, I think there needs to be a little more methodological explanation. Coming from the
outburst flood crowd rather than the environmental seismology crowd, I had to look up a lot
of terminology and there are abbreviations (STA/LTA) that go undefined. For an audience of
geomorphologists, I’d like to see a little more explanation of the methods and interpretation
of the waveforms, etc.

- I’m not sure I fully understand what new insights the clustering analysis brought to the table.
I see the authors have attributed to the clustering of events into collapse of the ice shelf, but
it’s unclear to me precisely why.

- (382) are the faults referred to here faults within the ice?

- There’s no figure 1

- (Fig. 3) “beamforming” another example of a term that might be well known to folks working
in seismology and signal processing but I haven’t come across as a geomorphologist. For ESurf
I think some of these terms just need clarification.

- (Fig 7c) lacks a scale

Referee Report: PDF

Hide

ED: Publish subject to minor revisions (review by editor) (17 Jun 2023) by Daniella Rempe

AR by Eva Eibl on behalf of the Authors (27 Jun 2023) Author's response Author's tracked changes Manuscript

ED: Publish as is (09 Jul 2023) by Daniella Rempe

ED: Publish as is (11 Jul 2023) by Tom Coulthard (Editor)

AR by Eva Eibl on behalf of the Authors (26 Jul 2023) Manuscript

Short summary

Floods draining beneath an ice cap are hazardous events that generate six different short- or long-lasting types of seismic signals. We use these signals to see the collapse of the ice once the water has left the lake, the propagation of the flood front to the terminus, hydrothermal explosions and boiling in the bedrock beneath the drained lake, and increased water flow at rapids in the glacial river. We can thus track the flood and assess the associated hazards better in future flooding events.