the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 04 Jun 2024
Dynamical changes of seismic properties prior to, during, and after 2014–2015 Holuhraun Eruption, Iceland
Abstract. In volcanic eruption monitoring, it is urgent to promptly detect changes in the volcanic system during the crisis period. Ideally continuous, temporally high-resolution, multidisciplinary data is available for this. However, some volcanoes are only being monitored using a single discipline or a single seismic station. In this case, it makes sense to harvest information from the available limited data set with several different techniques. Changes in the seismic complexity could reveal the dynamic changes due to magma propagation. We tested the performance of Permutation Entropy (PE) and Phase Permutation Entropy (PPE), which are fast and robust quantification of time series complexity, to monitor the change in the eruption process of 2014–2015 Holuhraun in Iceland. We additionally calculated the instantaneous frequency (IF), which is commonly used to monitor the frequency changes in a non-stationary signal. We observed distinct changes in the temporal variation of PE, PPE, and IF, which are consistent with the changing state from quiescence to magma propagation and then to eruption. During the eruption, PE and PPE fit the lava discharge rate, showing their potential to forecast the duration of the eruption. Finally, we also assessed the influence of the atmospheric noise to be considered in eruption monitoring.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(11678 KB)
-
Supplement
(2265 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(11678 KB) - Metadata XML
-
Supplement
(2265 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-1445', Anonymous Referee #1, 24 Jun 2024
This manuscript contains a very careful analysis of dynamical changes of continuous seismic waveforms recorded during the diking episode
and subsequent eruption of Bardarbunga volcano in 2014-2015. The analysis focuses on the use of Permutation Entropy and other useful
metrics that can be applied even if only one station is available which is a common state of affairs in under-monitored volcanoes. The
most interesting results of this work in my opinion are first, the good correlation of PE with discharge rate, and second, the great
sensitivity that PE exhibits in detecting minute changes in the seismic wavefield due to temperature variations. I think that the
manuscript is a good contribution to the field of volcano monitoring, it is clearly written, and it falls within the scope of the
journal. I therefore recommend publication after the comments listed below are taken into account by the authors.1. The part of the manuscript on clustering of the estimated parameters is the weakest since it is subjective and it does not take
into account clustering of other combinations of these parameters. Machine learning algorithms are widely used and software packages
that implement them are readily available, hence I don't understand why the authors have not attempted to perform an objective
analysis of clustering. If such algorithms are applied, then it will be possible to examine clustering beyond the binary (i.e two
parameters at a time) way presented in the manuscript, but rather consider clustering in a higher-dimensional space (e.g., PE, PPE,
IF, discharge rate etc).2. At the end of the conclusions section the authors state that "...our study shows that it is still challenging to determine the
eruption onset after the start of the pre-eruptive seismcity..."; I think that this statement may be accurate for the Holuhraun
diking-eruption case but not for other studied cases that utilized PE such as Gjalp or Shinmoedake cited by the authors. The main
difference between Holuhraun and these past cases is the intense volcanotectonic seismicity prior to eruption due to the lateral
propagation of the dike. The last part of the conclusions should be rephrased so that the reader becomes aware of this crucial
difference and this methodology may be able to pinpoint the eruption onset when seismicity levels are lower.3. In the supporting information Figures S3, and S5-S12 are plotted in a way that is difficult for the reader to easily follow the
description of the authors in the Discussion section. I would suggest that these Figures are redrafted using different color for
different lines by using a double y-axis, in the same way the authors plotted Figure 5 in the main manuscript.Citation: https://doi.org/10.5194/egusphere-2024-1445-RC1 -
AC1: 'Reply on RC1', Maria Sudibyo, 19 Aug 2024
Below, repetitions of the comments from the reviewer are written using regular font, and the response for the authors are written using italic font.
This manuscript contains a very careful analysis of dynamical changes of continuous seismic waveforms recorded during the diking episode and subsequent eruption of Bardarbunga volcano in 2014-2015. The analysis focuses on the use of Permutation Entropy and other useful metrics that can be applied even if only one station is available which is a common state of affairs in under-monitored volcanoes. The most interesting results of this work in my opinion are first, the good correlation of PE with discharge rate, and second, the great sensitivity that PE exhibits in detecting minute changes in the seismic wavefield due to temperature variations. I think that the manuscript is a good contribution to the field of volcano monitoring, it is clearly written, and it falls within the scope of the journal. I therefore recommend publication after the comments listed below are taken into account by the authors.
Response:We are very grateful to the reviewer for his positive statement and taking his time to provide useful feedback for the manuscript. Please find our responses to the specific comments below.
1. The part of the manuscript on clustering of the estimated parameters is the weakest since it is subjective and it does not take into account clustering of other combinations of these parameters. Machine learning algorithms are widely used and software packages that implement them are readily available, hence I don't understand why the authors have not attempted to perform an objective analysis of clustering. If such algorithms are applied, then it will be possible to examine clustering beyond the binary (i.e two parameters at a time) way presented in the manuscript, but rather consider clustering in a higher-dimensional space (e.g., PE, PPE, IF, discharge rate etc).
Response: We thank the reviewer for this suggestion. Initially, the aim of the ‘clustering’ section was simply to compare the expert interpretation of the ongoing volcanic activity with the spatial position (here 2-D projections) in the parameter space. We achieved this by coloring the different eruption stages given by the expert classification. We hoped to see that data points of PE, PPE, and mean IF fall into different groups associated with the different eruption stages. However, it is a good suggestion to use an objective clustering algorithm. Now we added a clustering analysis using K-Means with different combination of parameters, please see the results in the attachment. We also calculated the confusion matrix, which quantifies how many points in each cluster formed by K-Means lies in the equivalent clusters formed by the expert interpretation. We showed that using PE, PPE, mean IF, and log(RMeS) gives the highest score in the confusion matrix, therefore better in separating the eruption from the quiescence and the dyke propagation.
2. At the end of the conclusions section the authors state that "...our study shows that it is still challenging to determine the eruption onset after the start of the pre-eruptive seismcity..."; I think that this statement may be accurate for the Holuhraun diking-eruption case but not for other studied cases that utilized PE such as Gjalp or Shinmoedake cited by the authors. The main difference between Holuhraun and these past cases is the intense volcanotectonic seismicity prior to eruption due to the lateral propagation of the dike. The last part of the conclusions should be rephrased so that the reader becomes aware of this crucial difference and this methodology may be able to pinpoint the eruption onset when seismicity levels are lower.
Response: This is a good point and we have revised the conclusion section accordingly. In the new conclusion, we also noted that the parameters may respond differently to each stage of the eruption. While one parameter may be more sensitive to one stage, the other responds better to another stage, so combining them may provide more reliable information.
3. In the supporting information Figures S3, and S5-S12 are plotted in a way that is difficult for the reader to easily follow the description of the authors in the Discussion section. I would suggest that these Figures are redrafted using different color for different lines by using a double y-axis, in the same way the authors plotted Figure 5 in the main manuscript.
Response: Thank you for the feedback. We have revised the figures accordingly.
-
AC1: 'Reply on RC1', Maria Sudibyo, 19 Aug 2024
-
RC2: 'Comment on egusphere-2024-1445', Anonymous Referee #2, 30 Jun 2024
The manuscript “Dynamical changes of seismic properties prior to, during, and after 2014-2015 Holuhraun Eruption, Iceland” presents some interesting results regarding the application of Permutation and Phase Permutation Entropy to continuous seismic waveform timeseries prior, during and after the eruption process. Changes in these measures seem to be associated with the dynamic changes due to magma propagation and eruption, which are also pointed by changes in the instantaneous frequency. Remarkably, Permutation and Phase Permutation Entropy also correlate well with the lava discharge rate. Overall, the manuscript presents some interesting results regarding volcanic monitoring, is well structured and written and falls within the scope of NHESS. Therefore, I recommend its publication after revising some points.
1) I believe that section 5.5 “Cluster analysis” is the most weak in the manuscript. The authors claim that by separating the various measures into clusters according to the pre-eruptive and eruptive processes, then these measures show good separation during the various stages. This seems true during the dike propagation stages that clearly separate from the quiescence and eruptive periods. However, the last two show significant overlapping particularly when Permutation and Phase Permutation Entropy are plotted (Fig.6a). Please elaborate more on this issue.
2) Remove duplicate words from the text, as for instance in lines 27 and 408.
3) Although these are explained further in the text, introduce in line 78 what RMS, RMes and TADR stand for, as this is the first time that are encountered.
4) The colors for S1 and S2 in Fig.6 are not clearly distinctive.
5) Check your references, as doi for some seem to refer to other publications in the list.
Citation: https://doi.org/10.5194/egusphere-2024-1445-RC2 -
AC2: 'Reply on RC2', Maria Sudibyo, 19 Aug 2024
Below, repetitions of the comments from the reviewer are written using regular font and the response from the authors are written using italic font.
The manuscript “Dynamical changes of seismic properties prior to, during, and after 2014-2015 Holuhraun Eruption, Iceland” presents some interesting results regarding the application of Permutation and Phase Permutation Entropy to continuous seismic waveform timeseries prior, during and after the eruption process. Changes in these measures seem to be associated with the dynamic changes due to magma propagation and eruption, which are also pointed by changes in the instantaneous frequency. Remarkably, Permutation and Phase Permutation Entropy also correlate well with the lava discharge rate. Overall, the manuscript presents some interesting results regarding volcanic monitoring, is well structured and written and falls within the scope of NHESS. Therefore, I recommend its publication after revising some points.
Response: We are also very grateful to the second reviewer for providing helpful suggestions for the manuscript. Please find our responses to the specific comments below.
1) I believe that section 5.5 “Cluster analysis” is the most weak in the manuscript. The authors claim that by separating the various measures into clusters according to the pre-eruptive and eruptive processes, then these measures show good separation during the various stages. This seems true during the dike propagation stages that clearly separate from the quiescence and eruptive periods. However, the last two show significant overlapping particularly when Permutation and Phase Permutation Entropy are plotted (Fig.6a). Please elaborate more on this issue.
Response: We thank the reviewer for the critical feedback. The clustering figure presented in the original manuscript was done according to an expert interpretation and was plotted on 2D-planes; therefore, two clusters may appear to overlap. In response to the similar suggestion from Reviewer 1, we have now added the K-Means clustering 3D and 4D parametric space and compare them with the expert interpretation. We also now added a confusion matrix, to calculates how many points in each clusters formed by K-Means lies in the equivalent clusters by the expert interpretation. Please see the results in the attachment. We showed that using PE, PPE, mean IF and, log RMeS gives the highest score in the confusion matrix. Therefore better in separating the eruption from the quiescence and the dyke propagation. This part will be added to the revised section 5.5
2) Remove duplicate words from the text, as for instance in lines 27 and 408.
Response: We have rephrased those sentences to avoid duplicate words.
3) Although these are explained further in the text, introduce in line 78 what RMS, RMeS and TADR stand for, as this is the first time that are encountered.
Response: Thank you, we now introduced these parameters in the Introduction section
4) The colors for S1 and S2 in Fig.6 are not clearly distinctive.
Response: We have revised the figure using more distinctive colors that are still color-blind friendly.
5) Check your references, as doi for some seem to refer to other publications in the list.
Response: Thank you for your
careful and detail attention. We have checked and corrected the incorrect DOIs.
-
AC2: 'Reply on RC2', Maria Sudibyo, 19 Aug 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-1445', Anonymous Referee #1, 24 Jun 2024
This manuscript contains a very careful analysis of dynamical changes of continuous seismic waveforms recorded during the diking episode
and subsequent eruption of Bardarbunga volcano in 2014-2015. The analysis focuses on the use of Permutation Entropy and other useful
metrics that can be applied even if only one station is available which is a common state of affairs in under-monitored volcanoes. The
most interesting results of this work in my opinion are first, the good correlation of PE with discharge rate, and second, the great
sensitivity that PE exhibits in detecting minute changes in the seismic wavefield due to temperature variations. I think that the
manuscript is a good contribution to the field of volcano monitoring, it is clearly written, and it falls within the scope of the
journal. I therefore recommend publication after the comments listed below are taken into account by the authors.1. The part of the manuscript on clustering of the estimated parameters is the weakest since it is subjective and it does not take
into account clustering of other combinations of these parameters. Machine learning algorithms are widely used and software packages
that implement them are readily available, hence I don't understand why the authors have not attempted to perform an objective
analysis of clustering. If such algorithms are applied, then it will be possible to examine clustering beyond the binary (i.e two
parameters at a time) way presented in the manuscript, but rather consider clustering in a higher-dimensional space (e.g., PE, PPE,
IF, discharge rate etc).2. At the end of the conclusions section the authors state that "...our study shows that it is still challenging to determine the
eruption onset after the start of the pre-eruptive seismcity..."; I think that this statement may be accurate for the Holuhraun
diking-eruption case but not for other studied cases that utilized PE such as Gjalp or Shinmoedake cited by the authors. The main
difference between Holuhraun and these past cases is the intense volcanotectonic seismicity prior to eruption due to the lateral
propagation of the dike. The last part of the conclusions should be rephrased so that the reader becomes aware of this crucial
difference and this methodology may be able to pinpoint the eruption onset when seismicity levels are lower.3. In the supporting information Figures S3, and S5-S12 are plotted in a way that is difficult for the reader to easily follow the
description of the authors in the Discussion section. I would suggest that these Figures are redrafted using different color for
different lines by using a double y-axis, in the same way the authors plotted Figure 5 in the main manuscript.Citation: https://doi.org/10.5194/egusphere-2024-1445-RC1 -
AC1: 'Reply on RC1', Maria Sudibyo, 19 Aug 2024
Below, repetitions of the comments from the reviewer are written using regular font, and the response for the authors are written using italic font.
This manuscript contains a very careful analysis of dynamical changes of continuous seismic waveforms recorded during the diking episode and subsequent eruption of Bardarbunga volcano in 2014-2015. The analysis focuses on the use of Permutation Entropy and other useful metrics that can be applied even if only one station is available which is a common state of affairs in under-monitored volcanoes. The most interesting results of this work in my opinion are first, the good correlation of PE with discharge rate, and second, the great sensitivity that PE exhibits in detecting minute changes in the seismic wavefield due to temperature variations. I think that the manuscript is a good contribution to the field of volcano monitoring, it is clearly written, and it falls within the scope of the journal. I therefore recommend publication after the comments listed below are taken into account by the authors.
Response:We are very grateful to the reviewer for his positive statement and taking his time to provide useful feedback for the manuscript. Please find our responses to the specific comments below.
1. The part of the manuscript on clustering of the estimated parameters is the weakest since it is subjective and it does not take into account clustering of other combinations of these parameters. Machine learning algorithms are widely used and software packages that implement them are readily available, hence I don't understand why the authors have not attempted to perform an objective analysis of clustering. If such algorithms are applied, then it will be possible to examine clustering beyond the binary (i.e two parameters at a time) way presented in the manuscript, but rather consider clustering in a higher-dimensional space (e.g., PE, PPE, IF, discharge rate etc).
Response: We thank the reviewer for this suggestion. Initially, the aim of the ‘clustering’ section was simply to compare the expert interpretation of the ongoing volcanic activity with the spatial position (here 2-D projections) in the parameter space. We achieved this by coloring the different eruption stages given by the expert classification. We hoped to see that data points of PE, PPE, and mean IF fall into different groups associated with the different eruption stages. However, it is a good suggestion to use an objective clustering algorithm. Now we added a clustering analysis using K-Means with different combination of parameters, please see the results in the attachment. We also calculated the confusion matrix, which quantifies how many points in each cluster formed by K-Means lies in the equivalent clusters formed by the expert interpretation. We showed that using PE, PPE, mean IF, and log(RMeS) gives the highest score in the confusion matrix, therefore better in separating the eruption from the quiescence and the dyke propagation.
2. At the end of the conclusions section the authors state that "...our study shows that it is still challenging to determine the eruption onset after the start of the pre-eruptive seismcity..."; I think that this statement may be accurate for the Holuhraun diking-eruption case but not for other studied cases that utilized PE such as Gjalp or Shinmoedake cited by the authors. The main difference between Holuhraun and these past cases is the intense volcanotectonic seismicity prior to eruption due to the lateral propagation of the dike. The last part of the conclusions should be rephrased so that the reader becomes aware of this crucial difference and this methodology may be able to pinpoint the eruption onset when seismicity levels are lower.
Response: This is a good point and we have revised the conclusion section accordingly. In the new conclusion, we also noted that the parameters may respond differently to each stage of the eruption. While one parameter may be more sensitive to one stage, the other responds better to another stage, so combining them may provide more reliable information.
3. In the supporting information Figures S3, and S5-S12 are plotted in a way that is difficult for the reader to easily follow the description of the authors in the Discussion section. I would suggest that these Figures are redrafted using different color for different lines by using a double y-axis, in the same way the authors plotted Figure 5 in the main manuscript.
Response: Thank you for the feedback. We have revised the figures accordingly.
-
AC1: 'Reply on RC1', Maria Sudibyo, 19 Aug 2024
-
RC2: 'Comment on egusphere-2024-1445', Anonymous Referee #2, 30 Jun 2024
The manuscript “Dynamical changes of seismic properties prior to, during, and after 2014-2015 Holuhraun Eruption, Iceland” presents some interesting results regarding the application of Permutation and Phase Permutation Entropy to continuous seismic waveform timeseries prior, during and after the eruption process. Changes in these measures seem to be associated with the dynamic changes due to magma propagation and eruption, which are also pointed by changes in the instantaneous frequency. Remarkably, Permutation and Phase Permutation Entropy also correlate well with the lava discharge rate. Overall, the manuscript presents some interesting results regarding volcanic monitoring, is well structured and written and falls within the scope of NHESS. Therefore, I recommend its publication after revising some points.
1) I believe that section 5.5 “Cluster analysis” is the most weak in the manuscript. The authors claim that by separating the various measures into clusters according to the pre-eruptive and eruptive processes, then these measures show good separation during the various stages. This seems true during the dike propagation stages that clearly separate from the quiescence and eruptive periods. However, the last two show significant overlapping particularly when Permutation and Phase Permutation Entropy are plotted (Fig.6a). Please elaborate more on this issue.
2) Remove duplicate words from the text, as for instance in lines 27 and 408.
3) Although these are explained further in the text, introduce in line 78 what RMS, RMes and TADR stand for, as this is the first time that are encountered.
4) The colors for S1 and S2 in Fig.6 are not clearly distinctive.
5) Check your references, as doi for some seem to refer to other publications in the list.
Citation: https://doi.org/10.5194/egusphere-2024-1445-RC2 -
AC2: 'Reply on RC2', Maria Sudibyo, 19 Aug 2024
Below, repetitions of the comments from the reviewer are written using regular font and the response from the authors are written using italic font.
The manuscript “Dynamical changes of seismic properties prior to, during, and after 2014-2015 Holuhraun Eruption, Iceland” presents some interesting results regarding the application of Permutation and Phase Permutation Entropy to continuous seismic waveform timeseries prior, during and after the eruption process. Changes in these measures seem to be associated with the dynamic changes due to magma propagation and eruption, which are also pointed by changes in the instantaneous frequency. Remarkably, Permutation and Phase Permutation Entropy also correlate well with the lava discharge rate. Overall, the manuscript presents some interesting results regarding volcanic monitoring, is well structured and written and falls within the scope of NHESS. Therefore, I recommend its publication after revising some points.
Response: We are also very grateful to the second reviewer for providing helpful suggestions for the manuscript. Please find our responses to the specific comments below.
1) I believe that section 5.5 “Cluster analysis” is the most weak in the manuscript. The authors claim that by separating the various measures into clusters according to the pre-eruptive and eruptive processes, then these measures show good separation during the various stages. This seems true during the dike propagation stages that clearly separate from the quiescence and eruptive periods. However, the last two show significant overlapping particularly when Permutation and Phase Permutation Entropy are plotted (Fig.6a). Please elaborate more on this issue.
Response: We thank the reviewer for the critical feedback. The clustering figure presented in the original manuscript was done according to an expert interpretation and was plotted on 2D-planes; therefore, two clusters may appear to overlap. In response to the similar suggestion from Reviewer 1, we have now added the K-Means clustering 3D and 4D parametric space and compare them with the expert interpretation. We also now added a confusion matrix, to calculates how many points in each clusters formed by K-Means lies in the equivalent clusters by the expert interpretation. Please see the results in the attachment. We showed that using PE, PPE, mean IF and, log RMeS gives the highest score in the confusion matrix. Therefore better in separating the eruption from the quiescence and the dyke propagation. This part will be added to the revised section 5.5
2) Remove duplicate words from the text, as for instance in lines 27 and 408.
Response: We have rephrased those sentences to avoid duplicate words.
3) Although these are explained further in the text, introduce in line 78 what RMS, RMeS and TADR stand for, as this is the first time that are encountered.
Response: Thank you, we now introduced these parameters in the Introduction section
4) The colors for S1 and S2 in Fig.6 are not clearly distinctive.
Response: We have revised the figure using more distinctive colors that are still color-blind friendly.
5) Check your references, as doi for some seem to refer to other publications in the list.
Response: Thank you for your
careful and detail attention. We have checked and corrected the incorrect DOIs.
-
AC2: 'Reply on RC2', Maria Sudibyo, 19 Aug 2024
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 211 | 77 | 30 | 318 | 30 | 21 | 12 |
- HTML: 211
- PDF: 77
- XML: 30
- Total: 318
- Supplement: 30
- BibTeX: 21
- EndNote: 12
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Maria R.P. Sudibyo
Eva P. S. Eibl
Sebastian Hainzl
Matthias Ohrnberger
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(11678 KB) - Metadata XML
-
Supplement
(2265 KB) - BibTeX
- EndNote
- Final revised paper