the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 02 Jul 2026
Ice-core break-off as an opportunistic seismic source on the Northeast Greenland Ice Stream
Abstract. Ice-core drilling generates repeated mechanical disturbances within the ice column, raising the possibility that operational drilling signals could be used as opportunistic seismic sources. We test whether impulsive ice-core break-off events during the 2022 EastGRIP (Greenland) field season can be detected on surface-based, near-offset three-component geophones and used to estimate apparent P-wave velocity. A catalogue of 67 candidate core-break times was identified from drill-log load peaks, of which 37 had sufficient simultaneous near-offset seismic station coverage for assessment. Only three of these assessable events produced plausible impulsive arrivals. Automatic STA/LTA triggering was useful for identifying candidate windows, but did not reliably pick first arrivals. Where clear arrivals were identifiable above the background noise, manually picked arrivals gave apparent P-wave velocities consistent with firn/ice propagation, and showed a small across-flow faster-than-along-flow tendency, though this difference remains below the level of uncertainty required for robust anisotropy interpretation. These results show that ice-core break-off may provide a useful opportunistic seismic source, but only if future deployments include direct source-time measurement, quiet acquisition conditions, and receiver geometries designed for this purpose.
- Preprint
(1310 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-3762', Anonymous Referee #1, 21 Jul 2026
-
AC1: 'Reply on RC1', Emma Pearce, 07 Sep 2026
We thank the reviewer for their constructive review of the manuscript. We are pleased that the reviewer finds the concept interesting and recognises its value for future glaciological seismic-while-drilling experiments. LESSONS is a new article format for The Cryosphere, which started officially in May 2026, and provides a venue for reporting unsuccessful experiments and for sharing things that failed or did not work as planned. As this is one of the first submissions to LESSONS within The Cryosphere, we are keen to ensure that the revised manuscript fits the scope as clearly as possible. Thus, we are keen to ensure that the revised manuscript fits the scope as clearly as possible and know that by incorporating the comments from reviewer 1, this will be achieved.
Major comments
- I am afraid that the receiver geometry cannot resolve travel-time moveout. The source depth is approximately 2.2–2.4 km, whereas receiver offsets are only 10–100 m. For a source at 2300 m depth, the travel-path difference between 10 and 100 m offset is only about 2 m, corresponding to approximately 0.5 ms at 3800 m/s. This is much smaller than the 2.5 ms sampling interval of the 400 Hz data.
Therefore, the array cannot independently constrain velocity from differential arrival times. The reported apparent velocities are controlled primarily by the assumed source time rather than by measurable moveout. This limitation should be explicitly quantified.
We agree with the reviewer that the near-offset geometry cannot resolve travel-time moveout for a source at this depth. We will quantify this limitation in the manuscript and revise the manuscript so that the reported values are described as single effective apparent velocities through the firn-ice column.
- The tower-load maximum is only an indirect proxy for the actual fracture time, and the authors acknowledge that the true break may occur before, during, or after the load peak. The approximately 50 m/s difference between along-flow and across-flow groups corresponds to only several milliseconds of travel-time difference, which is likely smaller than the source-time and picking uncertainty. The directional comparison should therefore be removed or presented only as evidence that the current experiment is insufficient for anisotropy measurement. The calculated values should also be described as equivalent velocities conditional on the assumed source time, rather than independently measured P-wave velocities.
We agree that the tower-load maximum is only an indirect proxy for the true fracture time, and that uncertainty in source time is likely larger than the travel-time difference associated with the small along-flow/across-flow difference. We will therefore revise the manuscript so that the directional comparison is not presented as an anisotropy measurement but present it as an illustration of the limitation of the current experimental setup.
- The identified impulses may also originate from the drill tower, winch, drill cable, core barrel, camp activity, or other near-surface mechanical sources. Because the array provides almost no resolvable moveout, a deep source cannot currently be distinguished from a surface operational source.
The authors should use the three-component data to provide additional evidence, such as: polarization and incidence angle; waveform coherence and polarity across stations; amplitude variation with offset; comparison with surface-source and downhole-source hypotheses. Until this is demonstrated, the events should be called “candidate signals associated with core-break operations,” rather than confirmed core-break arrivals.
We agree that the identified impulses could potentially originate from other operational sources, and that the present array cannot independently distinguish a deep core-break source from a near-surface operational source using travel-time information alone.
In the revision, we will address this by using more cautious terminology throughout the manuscript where the events will be described as candidate signals associated with core-breaks. We will also make clear that the available geometry and source-time information do not support a robust source-location, polarisation, or incidence-angle analysis. This limitation will be explicitly included as one of the main lessons for future experimental design.
- The expected arrival window was defined using velocities between 2000 and 6000 m/s, and signals were then manually selected within that window. Any random impulse inside the window will necessarily produce a physically plausible apparent velocity. The authors should apply the same detection procedure to randomly shifted times or control windows without core breaks and estimate the false-positive rate. Ideally, manual classifications should be performed blindly, without displaying the predicted arrival window.
We agree that the broad search window increases the possibility that unrelated impulses could fall within the expected arrival window and appear physically plausible. The broad window was used deliberately to account for uncertainty in the true timing of core break-off relative to the tower-load maximum, but we recognise that this limits the strength of the interpretation. We will revise the manuscript to make this limitation clearer. We will also add a selection of control windows to assess the likelihood of false positives.
- Only three candidate detections and one non-detection are shown. For a Lessons Report, the unsuccessful cases are equally important. Please provide a table or summary figure for all assessable events, including: depth and load peak; available stations; background-noise level; detection confidence; signal-to-noise ratio; number of coherent stations. The authors should test whether detection depends on load, source depth, station coverage, or operational noise.
We agree that the unsuccessful cases are an important part of the contribution.
In the revision, we will add a clearer summary table of the full event set that will include depth, load peak, available stations and signal SNR.
- I am also thinking that the strongest contribution of this dataset is not a velocity or anisotropy measurement, but the identification of experimental requirements for future seismic-while-drilling studies. The manuscript would be more convincing if it focused on: detection probability; or source-time uncertainty; or geometric resolution; or required drill instrumentation. A known firn–ice VP(z) model should also be used to predict theoretical travel times. Observations could then be reported as timing residuals relative to the load maximum, rather than converted directly into apparent velocities.
We agree that the strongest contribution of the dataset is the identification of experimental requirements for future seismic-while-drilling studies in ice, and we will revise the manuscript to make this clearer. The purpose of publishing this dataset as a LESSONS Report is to show and discuss real-world data from seismic-while-drilling experiments, including what did not work as planned, so that future set-ups and acquisition strategies can be improved.
Minor comments
- Figure 3 should state which component is displayed. Three-component or rotated waveforms should be included for the three candidate detections.
We agree. We will revise the Figure 3 caption to state clearly which component is displayed.
- Picks from the N, E, and Z components of the same station are not independent observations. Figure 4 should preferably show one arrival estimate per station with an uncertainty.
We agree. In revision, Figure 4 will be changed so that it does not treat component picks from the same station as independent observations.
- The 20–200 Hz filter requires clarification because 200 Hz is the Nyquist frequency for data sampled at 400 Hz. Please report the actual filter corners, order, and implementation.
We agree. We will clarify the filter corners, order, and implementation, and ensure that the manuscript accurately describes the filtering applied to the 400~Hz data.
- Raw or less strongly filtered records should be provided to show that the identified impulses are not filter-induced ringing.
We agree that this would help demonstrate that the candidate impulses are not artefacts of filtering. We will clarify that the candidate impulses were inspected before and after filtering.
- Correct “propogate” to “propagate” and “ambiguious” to “ambiguous.”
We will correct these typographical errors.
Citation: https://doi.org/10.5194/egusphere-2026-3762-AC1
-
AC1: 'Reply on RC1', Emma Pearce, 07 Sep 2026
-
RC2: 'Comment on egusphere-2026-3762', Anonymous Referee #2, 02 Sep 2026
This manuscript tests whether ice-core break-off events can serve as an opportunistic seismic source during deep ice-core drilling, using drill-log load peaks as source-time proxies and near-offset geophones from the 2022 EastGRIP field season. Of 67 candidate events, only three yielded interpretable P-wave arrivals, giving velocity estimates broadly consistent with expected firn/ice values. The authors frame this honestly as a feasibility/lessons-learned study rather than a full success, and use it to outline what future deployments would need to make the method robust.
I found this a valuable and refreshingly transparent contribution. My main concerns are around confidence in the three detected events given the low detection rate and unconstrained source location (and time), the presentation of per-component picks in Figure 4, and the weight given to the anisotropy result relative to its stated uncertainty, detailed below, along with some minor comments.
Major comments:
- The detection rate of the 67 candidate break-off events is very low, which makes it hard to judge how confident the authors can be that the arrivals identified in the three usable core breaks are indeed genuine core-break signals rather than something else. This is especially true since the velocity estimates rely on an assumed source location, and that location is not independently constrained. Would it be possible to independently constrain the source location using these arrivals, to give more confidence in the origin of the signal? I suspect this would only be feasible where coverage is sufficient across several detections, especially including the larger-offset stations — so I'm not sure it's practical here, but it seems worth considering.
- I don't fully understand why Figure 4 presents the arrivals from different components as separate measurements. The fact that they vary so much mainly seems to convey that the picks are highly uncertain at each station, rather than adding independent information. I would suggest showing one pick per station instead, colour-coded as needed.
- I think the anisotropy aspect is worth mentioning, but I would be more cautious about presenting it as a result, including in the abstract As it is mentioned more than once that the difference between across and along flow apparent velocities lies well below the actual threshold of certainty. More generally, I would shift the paper's emphasis toward the lessons learned for improving the experimental setup, so that future work could produce a more extensive and usable dataset from core-break events.
Minor comments:
- Line 18: I would change "would" to "could," since you cannot be certain this approach would provide information on all of the subjects mentioned.
- Line 68: The reference to Figure 2 here is a bit misleading, as the figure only shows a limited time window of candidate core breaks.
- Figure 2: The caption could state more clearly that this is an example illustrating how candidates are selected.
- Figure 3: Please clarify in the caption whether the amplitude of each trace is normalized individually or shown on a common scale. Also, the pick lines are quite thick, which makes it somewhat hard to see the underlying waveform and first arrival — perhaps make them thinner and/or semi-transparent so the waveform remains visible.
- Figure 4: As noted above, I would avoid showing individual components separately and instead present one velocity estimate per station.
- Line 96: Space missing after (Figure 3)
Citation: https://doi.org/10.5194/egusphere-2026-3762-RC2 -
AC2: 'Reply on RC2', Emma Pearce, 07 Sep 2026
We thank the reviewer for their careful and constructive comments. We are pleased that the reviewer finds the manuscript valuable as a feasibility and lessons-learned study.
As this is a new article format for The Cryosphere, we are keen to ensure that the revised manuscript fits the scope as clearly as possible. Therefore, in the revision, we will make the framing more cautious and focus more directly on limitations and the requirements for future deployments.
Major comments
The detection rate of the 67 candidate break-off events is very low, which makes it hard to judge how confident the authors can be that the arrivals identified in the three usable core breaks are indeed genuine core-break signals rather than something else. This is especially true since the velocity estimates rely on an assumed source location, and that location is not independently constrained. Would it be possible to independently constrain the source location using these arrivals, to give more confidence in the origin of the signal? I suspect this would only be feasible where coverage is sufficient across several detections, especially including the larger-offset stations — so I'm not sure it's practical here, but it seems worth considering.
We agree that the low number of candidate detections and the unconstrained source time and location limit how confidently the selected arrivals can be attributed to core break-off at the borehole bottom. As also raised by Reviewer 1, the present near-offset geometry cannot independently resolve source location or travel-time moveout from a source at 2.4 km depth. Therefore, in the revision, we will address this by using more cautious terminology throughout the manuscript. The selected events will be described as candidate signals rather than confirmed core-break arrivals. We will also make clear that the apparent velocities are calculated from the assumed source depth and load-maximum source-time proxy, and should not be interpreted as independently measured P-wave velocities.
I don't fully understand why Figure 4 presents the arrivals from different components as separate measurements. The fact that they vary so much mainly seems to convey that the picks are highly uncertain at each station, rather than adding independent information. I would suggest showing one pick per station instead, colour-coded as needed.
We agree. Picks from different components of the same station should not be treated as independent observations. In the revised manuscript, Figure 4 will show one pick per event/station, represented as a single effective apparent velocity through the firn–ice column, with an uncertainty estimate.
I think the anisotropy aspect is worth mentioning, but I would be more cautious about presenting it as a result, including in the abstract. As it is mentioned more than once that the difference between across and along flow apparent velocities lies well below the actual threshold of certainty. More generally, I would shift the paper's emphasis toward the lessons learned for improving the experimental setup, so that future work could produce a more extensive and usable dataset from core-break events.
We agree. The directional comparison should not be presented as an anisotropy result. In the revised manuscript, we will reduce the emphasis placed on this comparison, including in the abstract, and will make clear that the dataset is insufficient for robust anisotropy interpretation.
Minor comments
Line 18: I would change "would" to "could," since you cannot be certain this approach would provide information on all of the subjects mentioned.
We agree and will change “would” to “could”.
Line 68: The reference to Figure 2 here is a bit misleading, as the figure only shows a limited time window of candidate core breaks.
We agree. We will revise the text so that Figure 2 is clearly described as an example of the candidate selection procedure, rather than as a full display of all candidate core breaks.
Figure 2: The caption could state more clearly that this is an example illustrating how candidates are selected.
We agree and will revise the Figure 2 caption to make clear that it shows an example of the selection method.
Figure 3: Please clarify in the caption whether the amplitude of each trace is normalized individually or shown on a common scale. Also, the pick lines are quite thick, which makes it somewhat hard to see the underlying waveform and first arrival — perhaps make them thinner and/or semi-transparent so the waveform remains visible.
We agree. We will revise the Figure 3 caption to state whether the traces are individually normalised or plotted on a common scale. We will also adjust the pick-line style.
Figure 4: As noted above, I would avoid showing individual components separately and instead present one velocity estimate per station.
We agree. As described above, we will revise Figure 4 to show one pick per event/station.
Line 96: Space missing after (Figure 3)
We will correct this formatting issue.
We thank the reviewer again for their helpful comments.
Citation: https://doi.org/10.5194/egusphere-2026-3762-AC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 111 | 40 | 23 | 174 | 24 | 18 |
- HTML: 111
- PDF: 40
- XML: 23
- Total: 174
- BibTeX: 24
- EndNote: 18
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript explores an interesting idea: using ice-core break-off during deep drilling as an opportunistic downhole seismic source. The concept could be useful for future glaciological seismic-while-drilling experiments, and the authors appropriately acknowledge the low detection rate and substantial operational limitations. However, the present data do not yet demonstrate that the three selected impulsive signals originated from core break-off at the borehole bottom. More importantly, the acquisition geometry and uncertain source time do not support an independent P-wave velocity estimate or a meaningful directional comparison. I therefore recommend major revision.
Major comments
1. I am afraid that the receiver geometry cannot resolve travel-time moveout. The source depth is approximately 2.2–2.4 km, whereas receiver offsets are only 10–100 m. For a source at 2300 m depth, the travel-path difference between 10 and 100 m offset is only about 2 m, corresponding to approximately 0.5 ms at 3800 m/s. This is much smaller than the 2.5 ms sampling interval of the 400 Hz data.
Therefore, the array cannot independently constrain velocity from differential arrival times. The reported apparent velocities are controlled primarily by the assumed source time rather than by measurable moveout. This limitation should be explicitly quantified.
2. The tower-load maximum is only an indirect proxy for the actual fracture time, and the authors acknowledge that the true break may occur before, during, or after the load peak. The approximately 50 m/s difference between along-flow and across-flow groups corresponds to only several milliseconds of travel-time difference, which is likely smaller than the source-time and picking uncertainty. The directional comparison should therefore be removed or presented only as evidence that the current experiment is insufficient for anisotropy measurement. The calculated values should also be described as equivalent velocities conditional on the assumed source time, rather than independently measured P-wave velocities.
3. The identified impulses may also originate from the drill tower, winch, drill cable, core barrel, camp activity, or other near-surface mechanical sources. Because the array provides almost no resolvable moveout, a deep source cannot currently be distinguished from a surface operational source.
The authors should use the three-component data to provide additional evidence, such as: polarization and incidence angle;
waveform coherence and polarity across stations; amplitude variation with offset; comparison with surface-source and downhole-source hypotheses. Until this is demonstrated, the events should be called “candidate signals associated with core-break operations,” rather than confirmed core-break arrivals.
4. The expected arrival window was defined using velocities between 2000 and 6000 m/s, and signals were then manually selected within that window. Any random impulse inside the window will necessarily produce a physically plausible apparent velocity. The authors should apply the same detection procedure to randomly shifted times or control windows without core breaks and estimate the false-positive rate. Ideally, manual classifications should be performed blindly, without displaying the predicted arrival window.
5. Only three candidate detections and one non-detection are shown. For a Lessons Report, the unsuccessful cases are equally important. Please provide a table or summary figure for all assessable events, including: depth and load peak;
available stations; background-noise level; detection confidence; signal-to-noise ratio; number of coherent stations. The authors should test whether detection depends on load, source depth, station coverage, or operational noise.
6. I am also thinking that the strongest contribution of this dataset is not a velocity or anisotropy measurement, but the identification of experimental requirements for future seismic-while-drilling studies. The manuscript would be more convincing if it focused on: detection probability; or source-time uncertainty; or geometric resolution; or required drill instrumentation. A known firn–ice VP(z) model should also be used to predict theoretical travel times. Observations could then be reported as timing residuals relative to the load maximum, rather than converted directly into apparent velocities.
Minor comments:
1. Figure 3 should state which component is displayed. Three-component or rotated waveforms should be included for the three candidate detections.
2. Picks from the N, E, and Z components of the same station are not independent observations. Figure 4 should preferably show one arrival estimate per station with an uncertainty.
3. The 20–200 Hz filter requires clarification because 200 Hz is the Nyquist frequency for data sampled at 400 Hz. Please report the actual filter corners, order, and implementation.
4. Raw or less strongly filtered records should be provided to show that the identified impulses are not filter-induced ringing.
5. Correct “propogate” to “propagate” and “ambiguious” to “ambiguous.”