the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
A template to obtain information on gravitational mass movements from the spectrograms of the seismic signals generated
Abstract. We present a template that helps in the classification of gravitational mass motions (snow avalanches, lahars and debris flows) by simply overlaying it graphically with the corresponding spectrogram of the generated seismic signal at the same scale. The template is created with different values of a parameter β' that allows us to analytically reproduce the exponential form of the increase in amplitude of high frequencies in time of the SON (Signal Onset) section of the spectrogram when the mass movement descends a slope and approach a seismic sensor. This increasing shape is a consequence of the appearance of energy at high frequencies as the gravitational mass approaches the seismic sensor. We present a methodology that includes a link between the propagation properties of seismic waves and the results of the application of an image processing using the Hough transform to demonstrate that this shape is related to the speed of the avalanche and the characteristics of the terrain. Seismic data generated by lahars, debris flows, and avalanches are used for the study. Depending on the type of event, differences are obtained in the order of magnitude of the values of β'. The mean value of β' for lahars is around 0.003 s-1, that for debris flows is an order of magnitude greater (0.017 s-1) and an order of magnitude less than that for avalanches (0.12 s-1). Furthermore, differences in β' are observed within each type of event.
Once the appropriated value of β' has been determined, the characteristics of the mass movement must be set according to expert judgement. This must be done for each site and for each type of gravitational mass movement. The application of the templates to the data of lahars and an avalanche recorded in two different places of its trajectory is presented as an example.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(1699 KB)
Interactive discussion
Status: closed
-
RC1: 'Comment on esurf-2022-10', Fabian Walter, 26 Oct 2022
The study by Suriñach and Flores-Márquez presents a spectral analysis of continuous seismic data containing the signature of lahars, debris flows and snow avalanches. The authors apply image processing techniques to the corresponding spectrograms in order to automatically recognize and characterize the well-known increase of high-frequency signal content as the mass movement approaches the recording station. With a simple analytical formula, it is possible to model the spectrogram envelopes and thus calculate the mass movement’s velocity.
The presented idea of coupling image processing with a physical understanding of frequency content of mass movement seismograms is clever and provides new perspectives for monitoring. Unfortunately, the manuscript is difficult to read, containing unclear as well as unnecessary explanations and numerous language and grammar mistakes. To be considered for publication, the entire text has to be overhauled and proof-read. Below I elaborate on this point of criticism and provide additional comments.
Fabian Walter.
GENERAL COMMENTS
This manuscript is unnecessarily long. It contains trivial and/or general explanations on spectrograms and computer code in addition to unreferenced assertions and announcements of what is about to be discussed. The corresponding pieces of text should be removed. Moreover, the introduction contains many examples of previous seismological studies of mass movements and their spectral characteristics. However, the investigations by Tsai et al. (2012), Lai et al. (2018) and Wenner et al. (2019), all of which are most relevant to the present study as they also document changes in high-frequency content as a function of source-station distances, are not discussed. It is important to state the underlying assumption adopted by the authors that stochastic superposition of instantaneous particle impacts with infinitely wide frequency spectra are the seismic source processes. Otherwise, a source signature in the spectral content would have to be included (Tsai et al., 2012). The introduction could be shortened to much less than 50% of its current length by discussing the above papers together with a few milestone and review papers. Previous papers should be discussed and cited based on their main point and citations should be complete rather than “examples” of what has been done and could be done. It is also not clear to me what the difference between the current submission and Suriñach et al. (2020) is. Is it simple the inclusion of mass movements other than snow avalanches? If so, then the current manuscript would not require more than 3-4 pages to get across its point. I urge the authors to strictly define the differences between the two studies to avoid accusations of self-plagiarism.
Whereas the authors focus on spectrogram envelopes it should be acknowledged that this is not the only way to characterize a mass movement seismogram. In particular, the authors should consider to couple the envelopes with amplitudes, which also increase as the moving mass approaches the recording station (the basis of amplitude source location mentioned in the introduction). Can the two methods be combined to give a better constraint on mass movement velocity?
The discussion of the analytical function (Equation 1 or a subsequent version thereof) is incomplete and confusing. First of all, what is the physical basis? There exist different kinds of waves for which attenuation laws can be formulated (e.g., surface waves or body waves). Which wave type is assumed? Depending on the answer it may or may not be legitimate to apply the same kind of equation to infrasound waves. This requires justification! Some more comments on Equation 1: why is t in the denominator primed? When fitting this equation to the spectrogram envelopes, I believe that it should be squared according to the definition of a power spectrum. See Stein and Wysession (2003), Equation 30 on Page 384: the power spectrum is proportional to the SQUARED norm of the Fourier components. Finally, it makes no sense to refer to a power spectral density (“frequency transect” as used by the authors) at a specific frequency (e.g., Figure 1f): by definition, this is a density residing between two frequencies.
Even though I believe I understand the application of the analytical expressions to the spectrogram envelope, I did not follow the author’s explanations. The explanations on Line 274ff make no sense to me (although this may be a fault on my side). As an alternative, I would suggest the following approach: One can rearrange Equation 1 such that the right-hand side contains the exponential and 1/sqrt(r) terms, only. The left hand side then contains the ratio A(t,f)/AS(f_s,t_s). To define an envelope, one can use a specific value of this ratio, say 80%, whose shape in the spectrogram follows the right-hand side (multiplied by some constant factors).
When analyzing the different events, it seems that the authors simply aim at determining the beta and K values. It would be more convincing to compare the velocities measured with the spectrogram method against independent measurements. At least for those sites hosting more than one event, this should be possible.
The avalanches from the Vallée de la Sionne are classified differently from the ones from Ryggfonn (powder, transitional, wet as opposed to dry-dense, dry-mixed). Why is this so? Do Norwegian avalanches differ from Swiss avalanches? What is a transitioning and mixing avalanche?
It seems that one important aspect of this work is the potential automatic classification of mass movements. The usefulness of this should be further discussed: for instance, what are the situations when a station records a lahar and a snow avalanche and one has to automatically discriminate between the two?
The fitting of the envelope in the Hough transform makes no sense to me, although this may be a problem on my side: how do the lines in Figure 5e constrain the beta value? I did not understand the corresponding explanations. How are the uncertainties determined?
In its current state the Conclusion is not a conclusion but actually a summary.
Avoid 1-sentence paragraphs.
There exist many undefined acronyms.
SPECIFIC COMMENTS
Lines 114-115: This sentence needs justification.
Line 206: Give the equation number in Aki and Richards (1980). Ideally use a more recent version of this textbook.
Equation 2: I thought that the quality factor is defined in a way that it does NOT depend on frequency.
Line 232: I do not know how to read the expressions for Q(f) and c(f). At least for the latter there is a dimensional error. Also, a frequency dependence of c implies dispersion, which should be commented on.
Lines 242ff: I have to admit that I did not follow the variable changes between r’, R and r. Why is this necessary? According to Line 246, r is negative, which does not seem to agree with Figure 2. Similarly, the sentence on Line 259 (constant exponential term) makes no sense to me.
Lines 268-270: Unclear.
Line 287: Specify “physical processes”.
Line 329: To argue for a constant beta^prime, is it not possible to simply assume a constant velocity?
Lines 392-393: This sentence is trivial. However, the exact coupling is more complicated than suggested since the source mechanisms generating seismic and infrasound waves are different.
Line 401: eigenvalue ï natural frequency
Lines 455-456: This agreement is not clear to me.
Lines 475-478: Define the “same considerations” and explain/justify the statement in the latter sentence.
Line 485: Define “shot point”.
Lines 488-490: This sentence is unclear to me.
Lines 509ff: I cannot confirm the shifting effect of K. Not sure I get the point of Figure 16. Perhaps this could be shown better?
FIGURES
It seems that some figures do not appear in the order that they are mentioned.
All figure panels need labels.
Figure 3: Superscript “-1” is missing in the bottom right panel.
Figure 4: What unit does the color scale have?
Figure 5: Y-axis and color bar labels are missing in some panels. I would remove “SON” since it appears in all panels. What are the dashed and solid lines?
REFERENCES
Lai, V. H., Tsai, V. C., Lamb, M. P., Ulizio, T. P., & Beer, A. R. (2018). The seismic signature of debris flows: Flow mechanics and early warning at Montecito, California. Geophysical Research Letters, 45(11), 5528-5535.
Tsai, V. C., Minchew, B., Lamb, M. P., & Ampuero, J. P. (2012). A physical model for seismic noise generation from sediment transport in rivers. Geophysical Research Letters, 39(2).
Suriñach, E., Flores-Márquez, E. L., Roig-Lafon, P., Furdada, G., & Tapia, M. (2020). Estimation of avalanche development and frontal velocities based on the spectrogram of the seismic signals generated at the Vallée de la Sionne Test Site. Geosciences, 10(3), 113.
Wenner, M., Walter, F., McArdell, B., & Farinotti, D. (2019). Deciphering debris-flow seismograms at Illgraben, Switzerland. Association of Environmental and Engineering Geologists; special publication 28.
Citation: https://doi.org/10.5194/esurf-2022-10-RC1 -
AC1: 'Reply on RC1', Emma Surinach, 11 Jan 2023
The study by Suriñach and Flores-Márquez presents a spectral analysis of continuous seismic data containing the signature of lahars, debris flows and snow avalanches. The authors apply image processing techniques to the corresponding spectrograms in order to automatically recognize and characterize the well-known increase of high-frequency signal content as the mass movement approaches the recording station. With a simple analytical formula, it is possible to model the spectrogram envelopes and thus calculate the mass movement’s velocity.
The presented idea of coupling image processing with a physical understanding of frequency content of mass movement seismograms is clever and provides new perspectives for monitoring. Unfortunately, the manuscript is difficult to read, containing unclear as well as unnecessary explanations and numerous language and grammar mistakes. To be considered for publication, the entire text has to be overhauled and proof- read. Below I elaborate on this point of criticism and provide additional comments.
Fabian Walter.
GENERAL COMMENTS
Thank you very much. This is a good opportunity to clarify concepts that you though there were know.
This manuscript is unnecessarily long. It contains trivial and/or general explanations on spectrograms and computer code in addition to unreferenced assertions and announcements of what is about to be discussed. The corresponding pieces of text should be removed.
You are right about long; we have removed some paragraphs.
The idea of introducing section 2 was to clarify to readers that there are differences between the spectrogram and the spectrum, and their understanding are fundamental for our purpose. We are dealing with frequency variations and not with amplitudes, as current studies usually do. However, we try to shorten as much as possible.
Moreover, the introduction contains many examples of previous seismological studies of mass movements and their spectral characteristics. However, the investigations by Tsai et al. (2012), Lai et al. (2018) and Wenner et al. (2019), all of which are most relevant to the present study as they also document changes in high-frequency content as a function of source-station distances, are not discussed. it is important to state the underlying assumption adopted by the authors that stochastic superposition of instantaneous particle impacts with infinitely wide frequency spectra are the seismic source processes. Otherwise, a source signature in the spectral content would have to be included (Tsai et al., 2012).
We are introduced the relevant papers you mentioned, that are theoretical studies dealing with the particle size related to the frequency content, but at the source. However, the propagation of seismic waves, intrinsic attenuation, the amplification of the site that affects the content of frequencies and amplitudes are not contemplated.
Because our templates are applied in the same location for comparison, the site features are the same and their effect can be avoided as they do not vary from site to site.
However, we include a sentence related to this point.
The introduction could be shortened to much less than 50% of its current length by discussing the above papers together with a few milestone and review papers.
Previous papers should be discussed and cited based on their main point and citations should be complete rather than “examples” of what has been done and could be done.
It is also not clear to me what the difference between the current submission and Suriñach et al. (2020) is. Is it simple the inclusion of mass movements other than snow avalanches? If so, then the current manuscript would not require more than 3-4 pages to get across its point. I urge the authors to strictly define the differences between the two studies to avoid accusations of self-plagiarism.
You are right, we have removed some paragraphs in the introduction, and it is centered in the goal of this new contribution. The goal of our document is to present a template for professionals to apply to the mass movement they are studying or detecting. While it is a tool, we demonstrate the principles of this too in section 3.
For more information see comments to reviewer 2 on this subject
Whereas the authors focus on spectrogram envelopes it should be acknowledged that this is not the only way to characterize a mass movement seismogram. In particular, the authors should consider to couple the envelopes with amplitudes, which also increase as the moving mass approaches the recording station (the basis of amplitude source location mentioned in the introduction). Can the two methods be combined to give a better constraint on mass movement velocity?
The amplitudes are a part of the wave field information. The frequency content is also important and we deal with this part. We propose this approach to simplify identification and provide professionals with a work tool.
Signal amplitude and frequency content are two independent variables of the wave field information. Increasing the amplitudes affects all frequencies equally. In addition, the effects of the ground can affect. Our approach deals also with frequencies, while other contributions mainly consider amplitudes. The amplification response of the ground can affects the distribution of the frequency content and must be taken into account.
The speed of advance of the beta parameters for different avalanches was discussed in Suriñach et al, (2020) but there were speed observations by different instruments to control the values. In case of doing the study, independent speed meters should be placed to see the dependence of the terrain factors. Once determined, then yes, speed estimates can be made.
The discussion of the analytical function (Equation 1 or a subsequent version thereof) is incomplete and confusing. First of all, what is the physical basis? There exist different kinds of waves for which attenuation laws can be formulated (e.g., surface waves or body waves). Which wave type is assumed? Depending on the answer it may or may not be legitimate to apply the same kind of equation to infrasound waves.
Essentially, we try to clarify in section 3 the theoretical reasoning for interpreting the exponential behavior of the shape in SON_spectrogram. In this new manuscript we have used the term SON_spectrogram ¨ instead of SON section of the spectrogram. This expression (eq. 1) represents the variation with distance in the amplitude, depending on the frequency, of a specific signal. This would be the variation of the signal observed at different hypothetical sensors along the path (variation in r'). This equation is derived from the variation of energy with distance. Note that the energy of a wave is proportional to the square of its amplitude.
The quotient term of this equation (the square root) explains the change in amplitude due to geometric spreading. We assumed surface waves, (it was already indicated in lines 212-213 of the previous manuscript) and the expression is related to the decrease in the energy flux at a point when the wave front propagates as a cylinder of radius r' and height h. The figure is found in Vilajosana et al., (2007) and in Suwa et al. (2003). Since the energy of a wave is proportional to the square of the amplitude, the square root appears.
The term AS is the amplitude of the source which is assumed not to be constant.
The exponential term explains the intrinsic attenuation which implies that the frequency content varies with distance. In the expression, the exponential is decreasing with increasing r'.
When fitting this equation to the spectrogram envelopes, I believe that it should be squared according to the definition of a power spectrum.
Note that the vertical axis of the spectrograms is Hz. If I understand correctly, it seems you are mixing amplitudes (PSD) and frequencies.
Thank you very much for the comment. Taking the squared in eq (4) we introduced this change to be more coherent, although the results we are interested in are the same. Here is the development although in the text the development is not presented.
In the pdf is the development
Even though I believe I understand the application of the analytical expressions to the spectrogram envelope, I did not follow the author’s explanations. The explanations on Line 274ff make no sense to me (although this may be a fault on my side). As an alternative, I would suggest the following approach: One can rearrange Equation 1 such that the right- hand side contains the exponential and 1/sqrt(r) terms, only. The left hand side then contains the ratio A(t,f)/AS(f_s,t_s). To define an envelope, one can use a specific value of this ratio, say 80%, whose shape in the spectrogram follows the right-hand side (multiplied by some constant factors).
Notice that your reasoning is with amplitudes and not with frequencies, we consider the changes in the frequency content.
We are not interested in the source itself. We are interested in the change of the spectrogram envelope with time and distance. In the case of establishing As = 1 the results would be the same.
Finally, it makes no sense to refer to a power spectral density (“frequency transect” as used by the authors) at a specific frequency (e.g., Figure 1f): by definition, this is a density residing between two frequencies.
To be more precise is the transect to the window f ± df. In our case for 9.3 ± 0.0156 Hz
When analyzing the different events, it seems that the authors simply aim at determining the beta and K values. It would be more convincing to compare the velocities measured with the spectrogram method against independent measurements. At least for those sites hosting more than one event, this should be possible.
You are right, but since the speeds depend on the characteristics of the terrain, it is necessary to adapt the equation to the characteristics of the soil, as was done in Suriñach et al. (2020).
The aim of this contribution is to provide a tool (the template) to professionals. Due to the different curves, the movements of the masses can be differentiated in each case. If professionals are interested in determining an estimate of the speed of mass movement, this contribution together with Suriñach et al., 2020, provide the necessary elements for further analysis. However, a previous characterization of the terrain is necessary.
The avalanches from the Vallée de la Sionne are classified differently from the ones from Ryggfonn (powder, transitional, wet as opposed to dry-dense, dry-mixed). Why is this so? Do Norwegian avalanches differ from Swiss avalanches
There is not a unified classification of snow avalanches according to their type. Avalanches depend on the temperature and the humidity of the snow and the ambient conditions (temperature, humidity...). Avalanches can evolve along their path. The terminology for its classification in this contribution is that used by the NGI and SLF groups. Even today, these two institutions use different terminology (e.g., Gauer et al, 2020, Sovilla et al., 2022). We prefer to keep the terminology used by each group. Moreover, depending on the latitude and altitude the avalanches can differ.
What is a transitioning and mixing avalanche?
A transitional avalanche is an avalanche that starts as a powder snow avalanche and due to different track altitude, snow incorporation, etc. evolves into a dense snow avalanche (e.g., Pérez-Guillén et al. 2016). A mixed avalanche does not imply evolution along the path.
It seems that one important aspect of this work is the potential automatic classification of mass movements. The usefulness of this should be further discussed: for instance, what are the situations when a station records a lahar and a snow avalanche, and one has to automatically discriminate between the two?
As already mentioned, the goal is to provide a tool for experts and professionals and not the automatic classification. The term classification is always introduced with the term help before it, and on line 114 it is enclosed in quotation marks.
Professionals are aware of the type of mass movements to be expected in each specific location. Also depending on the times of the year, one type of mass movement or another is expected. Also, the time and space scales of lahars and snow avalanches are different. Table 6 shows the related values and shows the difference in the different parameters of length and K and beta. A clarification in the text has been added
The fitting of the envelope in the Hough transform makes no sense to me, although this may be a problem on my side: how do the lines in Figure 5e constrain the beta value? I did not understand the corresponding explanations. How are the uncertainties determined?
Since this work presents a template for professionals to apply to the mass movement, we removed the part corresponding to Hough transform, as it is a tool to obtain the beta values and it is explained in Suriñach et al 2020. Please note that in the previous manuscript eq. (5) is exponential and if logarithms are applied it becomes a linear equation (9). The parameters of the line obtained through the Hough transform are those of equation (9). The uncertainties derive from the fitting of the different lines. Since the spectrogram is normally used as an image, we have used an image processing method to quantify the spectrogram image.
in its current state the Conclusion is not a conclusion but actually a summary.
We have rewritten discussion and conclusion sections
Avoid 1-sentence paragraphs.
Done
There exist many undefined acronyms.
Done, if you are referring to PSD and STA/LTA…. Moreover, we have added a list of acronymous at the end of the paper.
SPECIFIC COMMENTS
Lines 114-115: This sentence needs justification.
The increasing shape is due to the attenuation, and not to the amplitudes. e.g. By doubling the amplitude of the mass movement, the entire spectrum will increase by a factor of 4, but not the relationship between the frequency content, that is, the shape of the envelope. See eq. (1) and propagate accordingly.
Line 206: Give the equation number in Aki and Richards (1980). Ideally use a more recent version of this textbook.
This is not a direct equation. It is obtained by the combination of the content of subsection 5.5 and Box (5.7 eq. 4) of Aki and Richards (2002). Also, the content of subsections 3.7.5 in and 3.7.6 of Stein and Wysession (2003). The expression considers the damping by intrinsic attenuation and geometrical spreading.
Equation 2: I thought that the quality factor is defined in a way that it does NOT depend on frequency.
The quality factor Q depends on the frequency; however, it is a dimensionless variable because it is obtained from a ratio of amplitudes (for example, Stein and Wyssession, (2003) ch. 3.7 eq. 36). Also, the values of Q and its dependence on f (or w) vary for different locations. In e.g., (Levy et al., 2015, Appendix A) different values of Q are discussed.
In addition, the dependence of Q(f) and c(f) on the results of the energy calculation is evident (for example, formula (3) of the aforementioned work or in Suriñach et al., (2018).
In some cases, Q(f) is assumed to be constant over a specific frequency range (e.g., 5 Hz). In our case, since the frequency spectrum is broader, in principle, it must be considered.
As mentioned in the manuscript, the analytical expression was calculated in Vilajosana et al., (2007a) from experiments in Ryggfonn, this is not the case in our study since these parameters are collapsed in 𝛽 and as the template is for each location and for comparison purposes, in principle they should be the same.
Line 232: I do not know how to read the expressions for Q(f) and c(f). At least for the latter there is a dimensional error. Also, a frequency dependence of c implies dispersion, which should be commented on.
Thank you very much. There is a typo error in Q(f). It must be Q(f) = 2.8 f^0.57 as indicated in Vilajosana et al., (2007a).
c(f) is the phase velocity in m/s, and the values are also those obtained in Vilajosana et al., (2007a). For c(f) = 8.9 f + 722 in m/s the dimensions are c(f) =af+b; [a]= L; [f]= T-1;[b]= LT-1
We use these values for the example. These values must be fixed in each case. These parameters are collapsed in the 𝛽 and as the template is for each site, they must be the same, in principle.
Lines 242ff: I have to admit that I did not follow the variable changes between r’, R and r. Why is this necessary? According to Line 246, r is negative, which does not seem to agree with Figure 2. Similarly, the sentence on Line 259 (constant exponential term) makes no sense to me.
The change of variable corresponds to a change in the coordinate origin. r' is the increasing distance as the wave (not the source) moves away from the source: the minimum distance equals 0, the maximum equals R. On the other hand, for r, the distance decreases: maximum distance R, minimum 0. In the first case the exponential is increasing and in the other case it is decreasing. As the observation is made at the sensor, the distance from the source to the sensor decreases. And for the change of variable to be maintained, the signs must be like this. R is constant for each case since it does not vary neither in time nor in distance. For each point source is the maximum distance from the source to the sensor. We have eliminated the figure as suggested by reviewer 2
Lines 268-270: Unclear.
The following sentence is incorporated: This is a consequence of Parseval's theorem (e.g., Brigham, (1974), eq. 4-19). The interpretation of this form of the theorem is that the total energy of a signal is the same in both the time and frequency domains. And the energy is proportional to the square of the amplitude. A clarification has been added in the text
Line 287: Specify “physical processes”.
Done e.g., generation of seismic energy by impacts and friction on the ground. Speed of the mass movement and mass incorporation.
e.g. Kanamori and Given, (1982).
Line 329: To argue for a constant beta^prime, is it not possible to simply assume a constant velocity?
The frequency dependence of 𝛽 ' through α in equation (7 of the previous manuscript) must be taken into account. To be strict, the role of alpha is important. Note that we are dealing with the variation of the frequency content in time. In principle, 𝛽 is not constant because it depends on alpha and the velocity. Since we are obtaining averages (Suriñach et al 2020), we can assume constant 𝛽.Lines 392-393: This sentence is trivial. However, the exact coupling is more complicated than suggested since the source mechanisms generating seismic and infrasound waves are different.
Please note that we are considering the infrasound signal generated by the seismic signal by coupling at the ground surface and not the infrasound generated by fluctuations in the air. In Ichihara et al. (2012) (Supplementary material, based on Landau and Lifshitz (1987)) the explanation can be followed. Since the sensor is limited to a high pass of 4.5 Hz, it does not allow recording of low frequencies. Properties of infrasound propagation and attenuation can be found, for example, in Kogelnig et al., 2014. A comment has been added in the text.
Line 401: eigenvalue ïƒ natural frequency
Ok. Changed by natural frequency
Lines 455-456: This agreement is not clear to me.
It was explained in the text (457-460). However, we have included a paragraph in the text. Because higher frequencies attenuate with distance more quickly than lower frequencies, wet avalanche signals will begin to be detected at closer distances to the sensor than other types of avalanches, or equivalently take longer to detect. We have changed the word results to observations.
Lines 475-478: Define the “same considerations” and explain/justify the statement in the latter sentence.
That is to say: consider that the transmission of waves in the ground is instantaneous because the speed of seismic waves is two orders of magnitude greater than the speed of avalanches.
In order not to repeat, we have added a short sentence.
To obtain these results we have considered the difference in propagation speeds of seismic signals and avalanches in a similar way to the case of VDLS avalanches.
Line 485: Define “shot point”.
The avalanches were triggered by explosives. A clarification is added in line 420.
Shot point is replaced by shoot point, the point where is the explosion.
Lines 488-490: This sentence is unclear to me.
it was clarified.
Lines 509ff: I cannot confirm the shifting effect of K. Not sure I get the point of Figure 16. Perhaps this could be shown better?
We have changed the figure. The two figures have been joined and arrows indicated the shifting have been added. Also, an explanation in the text
FIGURES
It seems that some figures do not appear in the order that they are mentioned.
To fix this we have rearranged the presentation of the events.
All figure panels need labels.
Figure 3: Superscript “-1” is missing in the bottom right panel.
Thank you. Corrected
Figure 4: What unit does the color scale have?
We included in all figures (m s-1)2 s. We have changed the dB (without units because it is a relative value) and indicated that the values are in log10 .
Figure 5: Y-axis and color bar labels are missing in some panels. I would remove “SON” since it appears in all panels. What are the dashed and solid lines?
They are the fit of the exponential function of the SON shape. However, this figure has been removed because it is found in Suriñach et al., 2020.
REFERENCES
Lai, V. H., Tsai, V. C., Lamb, M. P., Ulizio, T. P., & Beer, A. R. (2018). The seismic signature of debris flows: Flow mechanics and early warning at Montecito, California. (11), 5528-5535.
Tsai, V. C., Minchew, B., Lamb, M. P., & Ampuero, J. P. (2012). A physical model for seismic noise generation from sediment transport in rivers. 3 (2).
Suriñach, E., Flores-Márquez, E. L., Roig-Lafon, P., Furdada, G., & Tapia, M. (2020). Estimation of avalanche development and frontal velocities based on the spectrogram of the seismic signals generated at the Vallée de la Sionne Test Site.(3), 113.
Wenner, M., Walter, F., McArdell, B., & Farinotti, D. (2019). Deciphering debris-flow seismograms at Illgraben, Switzerland.
REFERENCES
Aki, K. and Richards P.G. (2002). Quantitative Seismology (2on Edition). University Science Books.
Brigham, E.O. (1974). The fast Fourier transform. Pearson, 2014
Gauer et al, 2020, https://doi.org/10.1016/j.coldregions.2020.103165
Kanamori and Given., (1982) Analysis of long-period seismic waves excited by the May 1980, eruption of Mount St. Helens- A terrestrial monopole? J. Geophys. Res., 87, 5422-5432
Ichihara, M., Takeo, M., Yokoo, A., Oikawa, J., and Ohminato T. (2012). Monitoring volcanic activity using correlation patterns between infrasound and ground motion, Geophys. Res. Lett., 39, L04304, doi:10.1029/2011GL050542.
Landau, L.D- Lifshift (1987) Fluid mechanics, 2nd ed. Ed. Pergamon press, Oxford
Levy, C.; Mangeney, A.; Bonilla, F.; Hibert, C.; Calder, E.S.; Smith, P.J. (2015) Friction weakening in granular flows deduced from seismic records at the Soufrière Hills Volcano, Montserrat. J. Geophys. Res. Solid Earth, 120, 7536–7557, doi:10.1002/2015JB012151.
Pérez-Guillén, C.; Sovilla, B.; Suriñach, E.; Tapia, M.; Köhler, A.(2016). Deducing avalanche size and flow regimes from seismic measurements. Cold Reg. Sci. Technol. 2016, 121, 25–41, doi:10.1016/j.coldregions.2015.10.004.
Sovilla, B., Köhler, A., Kyburz, M., & Ligneau, C. (2019). The avalanche flow regimes and their pressure on infrastructures. In International Symposium on Mitigative Measures against Snow Avalanches and Other Rapid Gravity Mass Flows (pp. 223-228). Association of Chartered Engineers in Iceland. https://www.dora.lib4ri.ch/wsl/islandora/object/wsl:21560.
Stein, S., Wysession, M., 2003. An Introduction to Seismology, Earthquakes and Earth Structure. Blackwell Publishing.
Suwa, H., Akamatsu, J., Nagai, Y., 2003. Energy radiation by elastic waves from debris flows. Debris Flows Hazard Mitigation: Mechanics, Prediction and Assesement. Millpress, Rotterdam.
Suriñach, E., Tapia, M., Roig, P., and Blach, X. (2018). On the effect of the ground seismic characteristics in the estimation of mass movements based on seismic observation. Geophys. Res. Abstracts, EGU General Assembly 2018,20, https://meetingorganizer.copernicus.org/EGU2018/EGU2018-8479.pdf, https://www.researchgate.net/publication/33128412
Vilajosana, I.; Khazaradze, G.; Surinach, E.; Lied, E.; Kristensen, K. (2007a) Snow avalanche speed determination using seismic methods. Cold Reg. Sci. Technol., 49, 2–10.
Vilajosana, I.; Suriñach, E.; Khazaradze, G.; Gauer, P. (2007 b) Snow avalanche energy estimation from seismic signal analysis. Cold Reg. Sci. Technol., 50, 72–85.
-
AC1: 'Reply on RC1', Emma Surinach, 11 Jan 2023
-
RC2: 'Comment on esurf-2022-10', Anonymous Referee #2, 17 Nov 2022
First of all, I want to kindly thank the authors for their submission of the manuscript entitled “A template to obtain information on gravitational mass movements from the spectrograms of the seismic signals generated” by Suriñach and Flores-Márquez to the journal of Earth Surface Dynamics.
The manuscript discusses the spectral analysis of mass movement signals, and how to extract specific signatures from spectra of seismic data by fitting an exponential to the envelope of the signal onset.
The analysis of the signal onset section is a smart and intuitive way for the classification of seismic events, yet I believe that the manuscript should be shortened significantly and focus on the concise description of the method and/or the given applications before it should be considered for publication. In the sections below, you can find specific comments and a more detailed arguments for my opinion.
One significant issue I see is the similarity between the presented manuscript and the publication by Suriñach et al., 2020 (geosciences). I believe it is important to elaborate on what is actually new in the presented manuscript over the existing publication.
General Comments
- As I understand it, the presented algorithm is fitting an exponential onto the envelope of the signal onset within the spectrogram, resulting in exponents specific for a given type of mass-movements. This clear summary of the algorithm is missing, and instead is often circumscribed with unnecessarily complex paragraphs. Therefore, I believe the manuscript can be shortened significantly by removing redundant sections.
- What is the benefit of using the “template” instead of a thresholding of the fitted exponent values? Is the templating not an unnecessary additional step?
- The introduction reads more like an overview of interesting literature in related fields rather than a focused introduction working towards the main message of the manuscript. I believe that the manuscript would benefit from a more concise and focused introduction.
- If I understand and interpret this correctly, the same method as introduced and used in Suriñach et al., 2020 is used in this manuscript. To me, this was not clear when reading the manuscript. If this is in fact the case, I believe the manuscript could be shortened even more significantly, since the algorithm is already introduced in the above-mentioned publication.
- Please make sure that all acronyms are introduced. For example, PSD (Line 61), STA / LTA (Line 56) are not introduced properly. Also, for algorithms like STA/LTA, it would be great to have a citation in case someone is not familiar with these algorithms.
- The manuscript contains quite a number of redundant statements and paragraphs that contain not much information. Examples are in the paragraphs after line 75 and 125, where “code” and a developed “algorithm” are mentioned, but no information about the content is provided. I believe by removing such redundant and empty paragraphs, the manuscript could be shortened significantly by making it more concise.
- The word “considered” is repeated over and over, even in cases where other words might be more suitable (e.g., data were “used” instead of “considered”, the SON section was “analyzed” instead of “considered").
Specific Comments
Abstract:
- Line 10: you mention a “template” that you simply “overlay” over the data. When only reading the abstract, to me it is quite difficult to grasp what is meant by that. It would be great to get a better intuition of the workflow here, such that the abstract is clearer without reading the entire manuscript.
- Line 14: The way this is formulated is, that if a gravitational mass approaches the sensor, it will start emitting high frequencies. Is this not a path effect that for larger distances, the higher frequencies attenuate faster and do not reach the sensor, hence, the higher frequencies are only measurable closer to the sensor?
- Line 18: The specific values of beta’ in the abstract without context (e.g., an equation, behavior (liner / exponential)) makes it more confusing for me to read. What does this mean? 0.003 / second of what? I would probably explain this more or leave the specific values out.
1 Introduction:
- Line 66: The spectrum is a representation of the time-series data that can be visualized, but it is not really only a “visual representation”?
- Line 75: This paragraph is quite generic and is in my opinion not really needed. Instead, I would briefly mention how you calculate the spectrograms in one sentence.
- Figure 1: Is c) really needed? It contains the same information as in b. Also, in b) the colorbar label is missing.
- Line 125 to 130: Here you redundantly describe that there is a code and an algorithm that was developed. This is rather generic and does not provide any information about the algorithm itself and contains no information.
2 Characteristics of the data used
- Instead of the information collected in text, it would be easier readable if the sensors, dataloggers and characteristics (including references) were in the form of a table.
- 1: This could be re-named to Data pre-processing?
3 Spectrogram vs spectrum
- I believe that this section is not really needed, since the concept of spectra and spectrograms can be referred to as common knowledge in the seismic and seismological community.
- Figure 2: is this Figure really needed?
- Figure 4: Please add colorbar labels. (Please add colorbar labels to all figures and subfigures where they are missing).
4 Spectrogram treatment (ST)
- This is basically the core of the manuscript, describing the algorithm that is used. I believe this chapter could be shortened and better visualized in the form of a simple flowchart.
5 Application
- Figures 7,9 and 11: In the earlier figures, the subfigures were labelled with letters. For consistency, please do this for these figures, too. Colorbar labels are missing.
- I believe that this chapter could also be significantly shortened. Since the manuscripts focus is on the “new method”, I am not convinced that all the experiments and sites need to be introduced in such high detail, but I believe that the summary in Table 6 contains already a lot of the relevant information.
6 Discussion
- I really like table 6 and Figure 15. They are the main take-home messages for me
7 The template
- I have some worries about this section. As I understand, the shape (exponent) of the exponential is dependent on the type of mass-movement, the local geology and morphology (subsurface parameters and incidence angles) and distance of the mass-movement. This makes all the examples in 5. And 6. Site- and event-specific fits of an exponential line to the SON section. I am not yet convinced that the generalization into a template based on these findings adds a substantial benefit. Would it not be better to apply the algorithm in 4. In continuous time-windows to find signal onsets within your data, then calculate the exponents and use these for a classification, rather than “overlying” an arbitrary template on your spectrograms? What is the benefit of this template? I do not really see the reason to do this. Instead of using the template, could not a thresholding of the estimated exponents be a better approach for a classification, since all three investigated events fall within different order of magnitude ranges?
8 Conclusions
- See comments about section 7
- This section does more summarize the entire publication rather than draw conclusions.
- Have you tried the same approach for earthquake recordings or other events ? Might these overlap with the exponents of the shown data ?
- What happens in the absence of an event?
Mentioned literature:
Suriñach, E.; Flores-Márquez, E.L.; Roig-Lafon, P.; Furdada, G.; Tapia, M. Estimation of Avalanche Development and Frontal Velocities Based on the Spectrogram of the Seismic Signals Generated at the Vallée de la Sionne Test Site. Geosciences 2020, 10, 113. https://doi.org/10.3390/geosciences10030113
Citation: https://doi.org/10.5194/esurf-2022-10-RC2 -
AC2: 'Reply on RC2', Emma Surinach, 11 Jan 2023
Dear anonymous reviewer
Thankyou very much for your constructive comments.
Below you will find our replies, but first of all, we want to stress that the template of the envelopes of the spectrograms of the signals we present is a tool for practitioners to be applied in the “classification” to their specific events they are concerned. In principle, avalanche practitioners are not involved in lahars or in other different events and vice versa. Our tool is not for academia, although it is necessary to demonstrate their construction. The objective of this contribution is twofold, one to "academically" demonstrate the validity of obtaining the template, and secondly, to present the template to practitioners.
We have shortened the manuscript as suggested. In the introduction we highlight the difference with our previous paper (Suriñach et al., 2020). Although the method to obtain the spectrogram envelop is the same, the aim of the paper was to study the evolution along the avalanche path of different types of snow avalanches using the envelopes of the spectrograms of the generated seismic signals. We obtained differences in the evolutions of the shape according to the different values of k and beta and we obtained estimation of the speed compared with speeds obtained from external GEODAR measurements. Additionally, we selected the parameter of attenuation coefficient of ground α in the estimation of the avalanche speeds. This is not the aim of the present contribution in which we explain that the increasing shape of the spectrograms is a characteristic independent of the type of mass: snow avalanches at different geographical locations, debris flows and lahars when they approach a seismic sensor. Moreover, we obtain that the values of the 𝜥 and 𝛽 parameters differ between the different types of mass movements (avalanches, lahars and debris flow) and also between them (e.g., avalanche types). Only it is necessary that the mass approaches to the sensor.
As I understand it, the presented algorithm is fitting an exponential onto the envelope of the signal onset within the spectrogram, resulting in exponents specific for a given type of mass-movements. This clear summary of the algorithm is missing, and instead is often circumscribed with unnecessarily complex paragraphs. Therefore, I believe the manuscript can be shortened significantly by removing
We have shortened the manuscript as suggested
What is the benefit of using the “template” instead of a thresholding of the fitted exponent values? Is the templating not an unnecessary additional step?
One of the goals of this contribution is to provide the template as a tool for (technicians) professionals. Once the spectrogram of the generated seismic signal is obtained, professionals can “classify” the event, by using the template. Previously, they must build the template (with different values of 𝛽) and calibrate it.
- The introduction reads more like an overview of interesting literature in related fields rather than a focused introduction working towards the main message of the manuscript. I believe that the manuscript would benefit from a more concise and focused introduction.
Thanks, we have removed some paragraphs in the introduction, and it is centered in the goal of the paper. The goal of our document is to present in a justified manner a template for professionals to apply to the mass movement they are studying or detecting.
If I understand and interpret this correctly, the same method as introduced and used in Suriñach et al., 2020 is used in this manuscript. To me, this was not clear when reading the manuscript. If this is in fact the case, I believe the manuscript could be shortened even more significantly, since the algorithm is already introduced in the abovementioned publication.
In the new version of the introduction, we highlight the difference with our previous paper. Although the method to obtain the spectrograms envelop is the same, the aim of the paper was different. It was to study the evolution along the avalanche path of different types of snow avalanches and in the current study is to present a template that is valid for different gravitational mass movements.
We have eliminated the description of the method and referring to the previous paper.
- Please make sure that all acronyms are introduced. For example, PSD (Line 61), STA / LTA (Line 56) are not introduced properly. Also, for algorithms like STA/LTA, it would be great to have a citation in case someone is not familiar with these algorithms.
PSD ( Power Spectral Density)
STA/LTA Short time average/ long time average (e.g., Allen R.V. 1978) and Vaezi and Vander Baan (2015)
However, we have eliminated this part into the introduction.
- The manuscript contains quite a number of redundant statements and paragraphs that contain not much information. Examples are in the paragraphs after line 75 and 125, where “code” and a developed “algorithm” are mentioned, but no information about the content is provided. I believe by removing such redundant and empty paragraphs, the manuscript could be shortened significantly by making it more concise.
Thanks. We have removed the phrases. However, in the acknowledgments we indicate that the codes were created by our group.
- The word “considered” is repeated over and over, even in cases where other words might be more suitable (e.g., data were “used” instead of “considered”, the SON section was “analyzed” instead of “considered").
Thank you we have replaced these words.
Specific Comments
Abstract:
Line 10: you mention a “template” that you simply “overlay” over the data. When only reading the abstract, to me it is quite difficult to grasp what is meant by that. It would be great to get a better intuition of the workflow here, such that the abstract is clearer without reading the entire manuscript.
- Thank you, Done.
Line 14: The way this is formulated is, that if a gravitational mass approaches the sensor, it will start emitting high frequencies. Is this not a path effect that for larger distances, the higher frequencies attenuate faster and do not reach the sensor, hence, the higher frequencies are only measurable closer to the sensor?
YES. Thank you, we have tried to explain it better
Line 18: The specific values of beta’ in the abstract without context (e.g., an equation, behavior (liner / exponential)) makes it more confusing for me to read. What does this mean? 0.003 / second of what? I would probably explain this more or leave the specific values out.
Ok. Thank you, we have tried to explain it better
1 Introduction:
Line 66: The spectrum is a representation of the time-series data that can be visualized, but it is not really only a “visual representation”?
Yes, you are right, but we use the spectrogram as it is normally represented, as an image. This image contains numerical information, and we are using these values. Note that we mention that the spectrogram is a visual representation of the matrix values of the spectra, and we are using these values.
Line 75: This paragraph is quite generic and is in my opinion not really needed. Instead, I would briefly mention how you calculate the spectrograms in one sentence. Figure 1: Is c) really needed? It contains the same information as in b. Also, in b) the colorbar label is missing.
Figure 1c is obtained with the same information as the spectrogram but represented in 3D waterfall. It was shown to indicate that the spectrogram is not the only manner to visualize the matrix values. However, we have eliminated figure 1c.
We included in the colorbar of all figures (m s-1)2 s. We have changed the dB (without units because it is a relative value) and indicated that the values are in log10 in the figure caption.
Line 125 to 130: Here you redundantly describe that there is a code and an algorithm that was developed. This is rather generic and does not provide any information about the algorithm itself and contains no information.
We have eliminated this paragraph.
2 Characteristics of the data used
Instead of the information collected in text, it would be easier readable if the sensors, dataloggers and characteristics (including references) were in the form of a table.
OK; Table I is incorporated in the text
1: This could be re-named to Data pre-processing?
It is not opportune. In any case it would be pre-analysis of data. Everything mentioned in this section (filtering, FFT, etc.) is data processing like all other processes (ST) before the analysis.
3 Spectrogram vs spectrum
I believe that this section is not really needed, since the concept of spectra and spectrograms can be referred to as common knowledge in the seismic and seismological community.
The seismological community is more familiar with spectrum than with the spectrograms, which are used, normally as the image output of a software package. In our opinion it is better to clarify the differences between them, in order to avoid misunderstandings in the wide audience the journal of Earth Surface Dynamics.
Figure 2: is this Figure really needed?
For us is not needed, because only indicates a change of the reference system but notice the comment of the first reviewer.
We eliminate this figure as suggested.
Figure 4: Please add colorbar labels. (Please add colorbar labels to all figures and subfigures where they are missing).
We included in the colorbar of all figures (m s-1)2 s. We have changed the dB (without units because it is a relative value) and indicated that the values are in log10 in the figure caption.
4 Spectrogram treatment (ST)
This is basically the core of the manuscript, describing the algorithm that is used. I believe this chapter could be shortened and better visualized in the form of a simple flowchart.
We have introduced the following flowchart to clarify the theoretical assumptions and the procedure. In the pdf.
In this new manuscript we have used the term “SON_spectrogram” instead of SON section of the spectrogram.
5 Application
Figures 7,9 and 11: In the earlier figures, the subfigures were labelled with letters. For consistency, please do this for these figures, too. Colorbar labels are missing.
OK done
I believe that this chapter could also be significantly shortened. Since the manuscripts focus is on the “new method”, I am not convinced that all the experiments and sites need to be introduced in such high detail, but I believe that the summary in Table 6 contains already a lot of the relevant information.
The purpose was demonstrated that the behavior of the SON_spectrogram generated from seismic signal generated by mass movements in different situations and sites is the same. This is one of our “findings”. Table 6 without any evidence could be not convincing. We think that for our treatment and values to be credible it is necessary to demonstrate how they have been obtained.
6 Discussion
I really like table 6 and Figure 15. They are the main take-home messages for me
Thank you.
7 The template
I have some worries about this section. As I understand, the shape (exponent) of the exponential is dependent on the type of mass-movement, the local geology and morphology (subsurface parameters and incidence angles) and distance of the mass movement. This makes all the examples in 5. And 6. Site- and event-specific fits of an exponential line to the SON section. I am not yet convinced that the generalization into a template based on these findings adds a substantial benefit. Would it not be better to apply the m in 4. In continuous time-windows to find signal onsets within your data, then calculate the exponents and use these for a classification, rather than “overlying” an arbitrary template on your spectrograms? What is the benefit of this template? I do not really see the reason to do this. Instead of using the template, could not a thresholding of the estimated exponents be a better approach for a classification, since all three investigated events fall within different order of magnitude ranges?
The algorithm 4 is with the amplitudes of the time series not considering frequencies. This is another approach. Note that the exponent corresponds to the increasing shape of the amplitude frequencies in the spectrogram and, not of the amplitudes. When regarding amplitudes intrinsic attenuation is not considered.
One of the purposes of the paper is to supply a template for practitioners. We do not pretend practitioners to adjust numerically their data and found the exponents values. Spectrograms is an output of commercial software easy to obtain. Once the professionals have created and calibrated the template, they can "classify" their event by overlaying and sliding the template over the spectrogram. This allows obtaining the value of beta for a quantitative classification of the type of gravitational mass movement considered.
8 Conclusions
See comments about section 7
This section does more summarize the entire publication rather than draw conclusions.
We have rewritten discussion and conclusion sections
Have you tried the same approach for earthquake recordings or other events? Might these overlap with the exponents of the shown data?
As mentioned in the introduction, the increasing shape of the spectrograms (SON_spectrogram) is due to the existing elastic energy source moving towards a sensor and a product of the attenuation properties of the terrain. This is a peculiarity of mass movements, not necessarily gravitational ones (for example, Almendros et al. 2002). We include a sentence in the discussion.
In the first paragraph of the discussion a comment about this concerns is included.
What happens in the absence of an event?
In this case, no organized signals exist, only noise (wind, anthropogenic …). This is also a characteristic to take into account for the detection. We include a sentence in the discussion.
references
Allen R.V. (1978) Automatic earthquake recognition and timing from single traces Bull. Seism. Soc. Am., 68 1521- 1532
Almendros, J., Ibáñez, J.M., Alguacil, G., and Del Pezzo, E.: Array detection of a moving source, Seismol. Res. Lett. 73, 2 153-165, 2002.
Vaezi, Y and Van der Baan (2015) Comparison of the STA/LTA and power spectral density methods for microseismic event detection . Geophys. Jour. Int. , 203,3,1896-1908. https://doi.org/10.1093/gji/ggv419
Interactive discussion
Status: closed
-
RC1: 'Comment on esurf-2022-10', Fabian Walter, 26 Oct 2022
The study by Suriñach and Flores-Márquez presents a spectral analysis of continuous seismic data containing the signature of lahars, debris flows and snow avalanches. The authors apply image processing techniques to the corresponding spectrograms in order to automatically recognize and characterize the well-known increase of high-frequency signal content as the mass movement approaches the recording station. With a simple analytical formula, it is possible to model the spectrogram envelopes and thus calculate the mass movement’s velocity.
The presented idea of coupling image processing with a physical understanding of frequency content of mass movement seismograms is clever and provides new perspectives for monitoring. Unfortunately, the manuscript is difficult to read, containing unclear as well as unnecessary explanations and numerous language and grammar mistakes. To be considered for publication, the entire text has to be overhauled and proof-read. Below I elaborate on this point of criticism and provide additional comments.
Fabian Walter.
GENERAL COMMENTS
This manuscript is unnecessarily long. It contains trivial and/or general explanations on spectrograms and computer code in addition to unreferenced assertions and announcements of what is about to be discussed. The corresponding pieces of text should be removed. Moreover, the introduction contains many examples of previous seismological studies of mass movements and their spectral characteristics. However, the investigations by Tsai et al. (2012), Lai et al. (2018) and Wenner et al. (2019), all of which are most relevant to the present study as they also document changes in high-frequency content as a function of source-station distances, are not discussed. It is important to state the underlying assumption adopted by the authors that stochastic superposition of instantaneous particle impacts with infinitely wide frequency spectra are the seismic source processes. Otherwise, a source signature in the spectral content would have to be included (Tsai et al., 2012). The introduction could be shortened to much less than 50% of its current length by discussing the above papers together with a few milestone and review papers. Previous papers should be discussed and cited based on their main point and citations should be complete rather than “examples” of what has been done and could be done. It is also not clear to me what the difference between the current submission and Suriñach et al. (2020) is. Is it simple the inclusion of mass movements other than snow avalanches? If so, then the current manuscript would not require more than 3-4 pages to get across its point. I urge the authors to strictly define the differences between the two studies to avoid accusations of self-plagiarism.
Whereas the authors focus on spectrogram envelopes it should be acknowledged that this is not the only way to characterize a mass movement seismogram. In particular, the authors should consider to couple the envelopes with amplitudes, which also increase as the moving mass approaches the recording station (the basis of amplitude source location mentioned in the introduction). Can the two methods be combined to give a better constraint on mass movement velocity?
The discussion of the analytical function (Equation 1 or a subsequent version thereof) is incomplete and confusing. First of all, what is the physical basis? There exist different kinds of waves for which attenuation laws can be formulated (e.g., surface waves or body waves). Which wave type is assumed? Depending on the answer it may or may not be legitimate to apply the same kind of equation to infrasound waves. This requires justification! Some more comments on Equation 1: why is t in the denominator primed? When fitting this equation to the spectrogram envelopes, I believe that it should be squared according to the definition of a power spectrum. See Stein and Wysession (2003), Equation 30 on Page 384: the power spectrum is proportional to the SQUARED norm of the Fourier components. Finally, it makes no sense to refer to a power spectral density (“frequency transect” as used by the authors) at a specific frequency (e.g., Figure 1f): by definition, this is a density residing between two frequencies.
Even though I believe I understand the application of the analytical expressions to the spectrogram envelope, I did not follow the author’s explanations. The explanations on Line 274ff make no sense to me (although this may be a fault on my side). As an alternative, I would suggest the following approach: One can rearrange Equation 1 such that the right-hand side contains the exponential and 1/sqrt(r) terms, only. The left hand side then contains the ratio A(t,f)/AS(f_s,t_s). To define an envelope, one can use a specific value of this ratio, say 80%, whose shape in the spectrogram follows the right-hand side (multiplied by some constant factors).
When analyzing the different events, it seems that the authors simply aim at determining the beta and K values. It would be more convincing to compare the velocities measured with the spectrogram method against independent measurements. At least for those sites hosting more than one event, this should be possible.
The avalanches from the Vallée de la Sionne are classified differently from the ones from Ryggfonn (powder, transitional, wet as opposed to dry-dense, dry-mixed). Why is this so? Do Norwegian avalanches differ from Swiss avalanches? What is a transitioning and mixing avalanche?
It seems that one important aspect of this work is the potential automatic classification of mass movements. The usefulness of this should be further discussed: for instance, what are the situations when a station records a lahar and a snow avalanche and one has to automatically discriminate between the two?
The fitting of the envelope in the Hough transform makes no sense to me, although this may be a problem on my side: how do the lines in Figure 5e constrain the beta value? I did not understand the corresponding explanations. How are the uncertainties determined?
In its current state the Conclusion is not a conclusion but actually a summary.
Avoid 1-sentence paragraphs.
There exist many undefined acronyms.
SPECIFIC COMMENTS
Lines 114-115: This sentence needs justification.
Line 206: Give the equation number in Aki and Richards (1980). Ideally use a more recent version of this textbook.
Equation 2: I thought that the quality factor is defined in a way that it does NOT depend on frequency.
Line 232: I do not know how to read the expressions for Q(f) and c(f). At least for the latter there is a dimensional error. Also, a frequency dependence of c implies dispersion, which should be commented on.
Lines 242ff: I have to admit that I did not follow the variable changes between r’, R and r. Why is this necessary? According to Line 246, r is negative, which does not seem to agree with Figure 2. Similarly, the sentence on Line 259 (constant exponential term) makes no sense to me.
Lines 268-270: Unclear.
Line 287: Specify “physical processes”.
Line 329: To argue for a constant beta^prime, is it not possible to simply assume a constant velocity?
Lines 392-393: This sentence is trivial. However, the exact coupling is more complicated than suggested since the source mechanisms generating seismic and infrasound waves are different.
Line 401: eigenvalue ï natural frequency
Lines 455-456: This agreement is not clear to me.
Lines 475-478: Define the “same considerations” and explain/justify the statement in the latter sentence.
Line 485: Define “shot point”.
Lines 488-490: This sentence is unclear to me.
Lines 509ff: I cannot confirm the shifting effect of K. Not sure I get the point of Figure 16. Perhaps this could be shown better?
FIGURES
It seems that some figures do not appear in the order that they are mentioned.
All figure panels need labels.
Figure 3: Superscript “-1” is missing in the bottom right panel.
Figure 4: What unit does the color scale have?
Figure 5: Y-axis and color bar labels are missing in some panels. I would remove “SON” since it appears in all panels. What are the dashed and solid lines?
REFERENCES
Lai, V. H., Tsai, V. C., Lamb, M. P., Ulizio, T. P., & Beer, A. R. (2018). The seismic signature of debris flows: Flow mechanics and early warning at Montecito, California. Geophysical Research Letters, 45(11), 5528-5535.
Tsai, V. C., Minchew, B., Lamb, M. P., & Ampuero, J. P. (2012). A physical model for seismic noise generation from sediment transport in rivers. Geophysical Research Letters, 39(2).
Suriñach, E., Flores-Márquez, E. L., Roig-Lafon, P., Furdada, G., & Tapia, M. (2020). Estimation of avalanche development and frontal velocities based on the spectrogram of the seismic signals generated at the Vallée de la Sionne Test Site. Geosciences, 10(3), 113.
Wenner, M., Walter, F., McArdell, B., & Farinotti, D. (2019). Deciphering debris-flow seismograms at Illgraben, Switzerland. Association of Environmental and Engineering Geologists; special publication 28.
Citation: https://doi.org/10.5194/esurf-2022-10-RC1 -
AC1: 'Reply on RC1', Emma Surinach, 11 Jan 2023
The study by Suriñach and Flores-Márquez presents a spectral analysis of continuous seismic data containing the signature of lahars, debris flows and snow avalanches. The authors apply image processing techniques to the corresponding spectrograms in order to automatically recognize and characterize the well-known increase of high-frequency signal content as the mass movement approaches the recording station. With a simple analytical formula, it is possible to model the spectrogram envelopes and thus calculate the mass movement’s velocity.
The presented idea of coupling image processing with a physical understanding of frequency content of mass movement seismograms is clever and provides new perspectives for monitoring. Unfortunately, the manuscript is difficult to read, containing unclear as well as unnecessary explanations and numerous language and grammar mistakes. To be considered for publication, the entire text has to be overhauled and proof- read. Below I elaborate on this point of criticism and provide additional comments.
Fabian Walter.
GENERAL COMMENTS
Thank you very much. This is a good opportunity to clarify concepts that you though there were know.
This manuscript is unnecessarily long. It contains trivial and/or general explanations on spectrograms and computer code in addition to unreferenced assertions and announcements of what is about to be discussed. The corresponding pieces of text should be removed.
You are right about long; we have removed some paragraphs.
The idea of introducing section 2 was to clarify to readers that there are differences between the spectrogram and the spectrum, and their understanding are fundamental for our purpose. We are dealing with frequency variations and not with amplitudes, as current studies usually do. However, we try to shorten as much as possible.
Moreover, the introduction contains many examples of previous seismological studies of mass movements and their spectral characteristics. However, the investigations by Tsai et al. (2012), Lai et al. (2018) and Wenner et al. (2019), all of which are most relevant to the present study as they also document changes in high-frequency content as a function of source-station distances, are not discussed. it is important to state the underlying assumption adopted by the authors that stochastic superposition of instantaneous particle impacts with infinitely wide frequency spectra are the seismic source processes. Otherwise, a source signature in the spectral content would have to be included (Tsai et al., 2012).
We are introduced the relevant papers you mentioned, that are theoretical studies dealing with the particle size related to the frequency content, but at the source. However, the propagation of seismic waves, intrinsic attenuation, the amplification of the site that affects the content of frequencies and amplitudes are not contemplated.
Because our templates are applied in the same location for comparison, the site features are the same and their effect can be avoided as they do not vary from site to site.
However, we include a sentence related to this point.
The introduction could be shortened to much less than 50% of its current length by discussing the above papers together with a few milestone and review papers.
Previous papers should be discussed and cited based on their main point and citations should be complete rather than “examples” of what has been done and could be done.
It is also not clear to me what the difference between the current submission and Suriñach et al. (2020) is. Is it simple the inclusion of mass movements other than snow avalanches? If so, then the current manuscript would not require more than 3-4 pages to get across its point. I urge the authors to strictly define the differences between the two studies to avoid accusations of self-plagiarism.
You are right, we have removed some paragraphs in the introduction, and it is centered in the goal of this new contribution. The goal of our document is to present a template for professionals to apply to the mass movement they are studying or detecting. While it is a tool, we demonstrate the principles of this too in section 3.
For more information see comments to reviewer 2 on this subject
Whereas the authors focus on spectrogram envelopes it should be acknowledged that this is not the only way to characterize a mass movement seismogram. In particular, the authors should consider to couple the envelopes with amplitudes, which also increase as the moving mass approaches the recording station (the basis of amplitude source location mentioned in the introduction). Can the two methods be combined to give a better constraint on mass movement velocity?
The amplitudes are a part of the wave field information. The frequency content is also important and we deal with this part. We propose this approach to simplify identification and provide professionals with a work tool.
Signal amplitude and frequency content are two independent variables of the wave field information. Increasing the amplitudes affects all frequencies equally. In addition, the effects of the ground can affect. Our approach deals also with frequencies, while other contributions mainly consider amplitudes. The amplification response of the ground can affects the distribution of the frequency content and must be taken into account.
The speed of advance of the beta parameters for different avalanches was discussed in Suriñach et al, (2020) but there were speed observations by different instruments to control the values. In case of doing the study, independent speed meters should be placed to see the dependence of the terrain factors. Once determined, then yes, speed estimates can be made.
The discussion of the analytical function (Equation 1 or a subsequent version thereof) is incomplete and confusing. First of all, what is the physical basis? There exist different kinds of waves for which attenuation laws can be formulated (e.g., surface waves or body waves). Which wave type is assumed? Depending on the answer it may or may not be legitimate to apply the same kind of equation to infrasound waves.
Essentially, we try to clarify in section 3 the theoretical reasoning for interpreting the exponential behavior of the shape in SON_spectrogram. In this new manuscript we have used the term SON_spectrogram ¨ instead of SON section of the spectrogram. This expression (eq. 1) represents the variation with distance in the amplitude, depending on the frequency, of a specific signal. This would be the variation of the signal observed at different hypothetical sensors along the path (variation in r'). This equation is derived from the variation of energy with distance. Note that the energy of a wave is proportional to the square of its amplitude.
The quotient term of this equation (the square root) explains the change in amplitude due to geometric spreading. We assumed surface waves, (it was already indicated in lines 212-213 of the previous manuscript) and the expression is related to the decrease in the energy flux at a point when the wave front propagates as a cylinder of radius r' and height h. The figure is found in Vilajosana et al., (2007) and in Suwa et al. (2003). Since the energy of a wave is proportional to the square of the amplitude, the square root appears.
The term AS is the amplitude of the source which is assumed not to be constant.
The exponential term explains the intrinsic attenuation which implies that the frequency content varies with distance. In the expression, the exponential is decreasing with increasing r'.
When fitting this equation to the spectrogram envelopes, I believe that it should be squared according to the definition of a power spectrum.
Note that the vertical axis of the spectrograms is Hz. If I understand correctly, it seems you are mixing amplitudes (PSD) and frequencies.
Thank you very much for the comment. Taking the squared in eq (4) we introduced this change to be more coherent, although the results we are interested in are the same. Here is the development although in the text the development is not presented.
In the pdf is the development
Even though I believe I understand the application of the analytical expressions to the spectrogram envelope, I did not follow the author’s explanations. The explanations on Line 274ff make no sense to me (although this may be a fault on my side). As an alternative, I would suggest the following approach: One can rearrange Equation 1 such that the right- hand side contains the exponential and 1/sqrt(r) terms, only. The left hand side then contains the ratio A(t,f)/AS(f_s,t_s). To define an envelope, one can use a specific value of this ratio, say 80%, whose shape in the spectrogram follows the right-hand side (multiplied by some constant factors).
Notice that your reasoning is with amplitudes and not with frequencies, we consider the changes in the frequency content.
We are not interested in the source itself. We are interested in the change of the spectrogram envelope with time and distance. In the case of establishing As = 1 the results would be the same.
Finally, it makes no sense to refer to a power spectral density (“frequency transect” as used by the authors) at a specific frequency (e.g., Figure 1f): by definition, this is a density residing between two frequencies.
To be more precise is the transect to the window f ± df. In our case for 9.3 ± 0.0156 Hz
When analyzing the different events, it seems that the authors simply aim at determining the beta and K values. It would be more convincing to compare the velocities measured with the spectrogram method against independent measurements. At least for those sites hosting more than one event, this should be possible.
You are right, but since the speeds depend on the characteristics of the terrain, it is necessary to adapt the equation to the characteristics of the soil, as was done in Suriñach et al. (2020).
The aim of this contribution is to provide a tool (the template) to professionals. Due to the different curves, the movements of the masses can be differentiated in each case. If professionals are interested in determining an estimate of the speed of mass movement, this contribution together with Suriñach et al., 2020, provide the necessary elements for further analysis. However, a previous characterization of the terrain is necessary.
The avalanches from the Vallée de la Sionne are classified differently from the ones from Ryggfonn (powder, transitional, wet as opposed to dry-dense, dry-mixed). Why is this so? Do Norwegian avalanches differ from Swiss avalanches
There is not a unified classification of snow avalanches according to their type. Avalanches depend on the temperature and the humidity of the snow and the ambient conditions (temperature, humidity...). Avalanches can evolve along their path. The terminology for its classification in this contribution is that used by the NGI and SLF groups. Even today, these two institutions use different terminology (e.g., Gauer et al, 2020, Sovilla et al., 2022). We prefer to keep the terminology used by each group. Moreover, depending on the latitude and altitude the avalanches can differ.
What is a transitioning and mixing avalanche?
A transitional avalanche is an avalanche that starts as a powder snow avalanche and due to different track altitude, snow incorporation, etc. evolves into a dense snow avalanche (e.g., Pérez-Guillén et al. 2016). A mixed avalanche does not imply evolution along the path.
It seems that one important aspect of this work is the potential automatic classification of mass movements. The usefulness of this should be further discussed: for instance, what are the situations when a station records a lahar and a snow avalanche, and one has to automatically discriminate between the two?
As already mentioned, the goal is to provide a tool for experts and professionals and not the automatic classification. The term classification is always introduced with the term help before it, and on line 114 it is enclosed in quotation marks.
Professionals are aware of the type of mass movements to be expected in each specific location. Also depending on the times of the year, one type of mass movement or another is expected. Also, the time and space scales of lahars and snow avalanches are different. Table 6 shows the related values and shows the difference in the different parameters of length and K and beta. A clarification in the text has been added
The fitting of the envelope in the Hough transform makes no sense to me, although this may be a problem on my side: how do the lines in Figure 5e constrain the beta value? I did not understand the corresponding explanations. How are the uncertainties determined?
Since this work presents a template for professionals to apply to the mass movement, we removed the part corresponding to Hough transform, as it is a tool to obtain the beta values and it is explained in Suriñach et al 2020. Please note that in the previous manuscript eq. (5) is exponential and if logarithms are applied it becomes a linear equation (9). The parameters of the line obtained through the Hough transform are those of equation (9). The uncertainties derive from the fitting of the different lines. Since the spectrogram is normally used as an image, we have used an image processing method to quantify the spectrogram image.
in its current state the Conclusion is not a conclusion but actually a summary.
We have rewritten discussion and conclusion sections
Avoid 1-sentence paragraphs.
Done
There exist many undefined acronyms.
Done, if you are referring to PSD and STA/LTA…. Moreover, we have added a list of acronymous at the end of the paper.
SPECIFIC COMMENTS
Lines 114-115: This sentence needs justification.
The increasing shape is due to the attenuation, and not to the amplitudes. e.g. By doubling the amplitude of the mass movement, the entire spectrum will increase by a factor of 4, but not the relationship between the frequency content, that is, the shape of the envelope. See eq. (1) and propagate accordingly.
Line 206: Give the equation number in Aki and Richards (1980). Ideally use a more recent version of this textbook.
This is not a direct equation. It is obtained by the combination of the content of subsection 5.5 and Box (5.7 eq. 4) of Aki and Richards (2002). Also, the content of subsections 3.7.5 in and 3.7.6 of Stein and Wysession (2003). The expression considers the damping by intrinsic attenuation and geometrical spreading.
Equation 2: I thought that the quality factor is defined in a way that it does NOT depend on frequency.
The quality factor Q depends on the frequency; however, it is a dimensionless variable because it is obtained from a ratio of amplitudes (for example, Stein and Wyssession, (2003) ch. 3.7 eq. 36). Also, the values of Q and its dependence on f (or w) vary for different locations. In e.g., (Levy et al., 2015, Appendix A) different values of Q are discussed.
In addition, the dependence of Q(f) and c(f) on the results of the energy calculation is evident (for example, formula (3) of the aforementioned work or in Suriñach et al., (2018).
In some cases, Q(f) is assumed to be constant over a specific frequency range (e.g., 5 Hz). In our case, since the frequency spectrum is broader, in principle, it must be considered.
As mentioned in the manuscript, the analytical expression was calculated in Vilajosana et al., (2007a) from experiments in Ryggfonn, this is not the case in our study since these parameters are collapsed in 𝛽 and as the template is for each location and for comparison purposes, in principle they should be the same.
Line 232: I do not know how to read the expressions for Q(f) and c(f). At least for the latter there is a dimensional error. Also, a frequency dependence of c implies dispersion, which should be commented on.
Thank you very much. There is a typo error in Q(f). It must be Q(f) = 2.8 f^0.57 as indicated in Vilajosana et al., (2007a).
c(f) is the phase velocity in m/s, and the values are also those obtained in Vilajosana et al., (2007a). For c(f) = 8.9 f + 722 in m/s the dimensions are c(f) =af+b; [a]= L; [f]= T-1;[b]= LT-1
We use these values for the example. These values must be fixed in each case. These parameters are collapsed in the 𝛽 and as the template is for each site, they must be the same, in principle.
Lines 242ff: I have to admit that I did not follow the variable changes between r’, R and r. Why is this necessary? According to Line 246, r is negative, which does not seem to agree with Figure 2. Similarly, the sentence on Line 259 (constant exponential term) makes no sense to me.
The change of variable corresponds to a change in the coordinate origin. r' is the increasing distance as the wave (not the source) moves away from the source: the minimum distance equals 0, the maximum equals R. On the other hand, for r, the distance decreases: maximum distance R, minimum 0. In the first case the exponential is increasing and in the other case it is decreasing. As the observation is made at the sensor, the distance from the source to the sensor decreases. And for the change of variable to be maintained, the signs must be like this. R is constant for each case since it does not vary neither in time nor in distance. For each point source is the maximum distance from the source to the sensor. We have eliminated the figure as suggested by reviewer 2
Lines 268-270: Unclear.
The following sentence is incorporated: This is a consequence of Parseval's theorem (e.g., Brigham, (1974), eq. 4-19). The interpretation of this form of the theorem is that the total energy of a signal is the same in both the time and frequency domains. And the energy is proportional to the square of the amplitude. A clarification has been added in the text
Line 287: Specify “physical processes”.
Done e.g., generation of seismic energy by impacts and friction on the ground. Speed of the mass movement and mass incorporation.
e.g. Kanamori and Given, (1982).
Line 329: To argue for a constant beta^prime, is it not possible to simply assume a constant velocity?
The frequency dependence of 𝛽 ' through α in equation (7 of the previous manuscript) must be taken into account. To be strict, the role of alpha is important. Note that we are dealing with the variation of the frequency content in time. In principle, 𝛽 is not constant because it depends on alpha and the velocity. Since we are obtaining averages (Suriñach et al 2020), we can assume constant 𝛽.Lines 392-393: This sentence is trivial. However, the exact coupling is more complicated than suggested since the source mechanisms generating seismic and infrasound waves are different.
Please note that we are considering the infrasound signal generated by the seismic signal by coupling at the ground surface and not the infrasound generated by fluctuations in the air. In Ichihara et al. (2012) (Supplementary material, based on Landau and Lifshitz (1987)) the explanation can be followed. Since the sensor is limited to a high pass of 4.5 Hz, it does not allow recording of low frequencies. Properties of infrasound propagation and attenuation can be found, for example, in Kogelnig et al., 2014. A comment has been added in the text.
Line 401: eigenvalue ïƒ natural frequency
Ok. Changed by natural frequency
Lines 455-456: This agreement is not clear to me.
It was explained in the text (457-460). However, we have included a paragraph in the text. Because higher frequencies attenuate with distance more quickly than lower frequencies, wet avalanche signals will begin to be detected at closer distances to the sensor than other types of avalanches, or equivalently take longer to detect. We have changed the word results to observations.
Lines 475-478: Define the “same considerations” and explain/justify the statement in the latter sentence.
That is to say: consider that the transmission of waves in the ground is instantaneous because the speed of seismic waves is two orders of magnitude greater than the speed of avalanches.
In order not to repeat, we have added a short sentence.
To obtain these results we have considered the difference in propagation speeds of seismic signals and avalanches in a similar way to the case of VDLS avalanches.
Line 485: Define “shot point”.
The avalanches were triggered by explosives. A clarification is added in line 420.
Shot point is replaced by shoot point, the point where is the explosion.
Lines 488-490: This sentence is unclear to me.
it was clarified.
Lines 509ff: I cannot confirm the shifting effect of K. Not sure I get the point of Figure 16. Perhaps this could be shown better?
We have changed the figure. The two figures have been joined and arrows indicated the shifting have been added. Also, an explanation in the text
FIGURES
It seems that some figures do not appear in the order that they are mentioned.
To fix this we have rearranged the presentation of the events.
All figure panels need labels.
Figure 3: Superscript “-1” is missing in the bottom right panel.
Thank you. Corrected
Figure 4: What unit does the color scale have?
We included in all figures (m s-1)2 s. We have changed the dB (without units because it is a relative value) and indicated that the values are in log10 .
Figure 5: Y-axis and color bar labels are missing in some panels. I would remove “SON” since it appears in all panels. What are the dashed and solid lines?
They are the fit of the exponential function of the SON shape. However, this figure has been removed because it is found in Suriñach et al., 2020.
REFERENCES
Lai, V. H., Tsai, V. C., Lamb, M. P., Ulizio, T. P., & Beer, A. R. (2018). The seismic signature of debris flows: Flow mechanics and early warning at Montecito, California. (11), 5528-5535.
Tsai, V. C., Minchew, B., Lamb, M. P., & Ampuero, J. P. (2012). A physical model for seismic noise generation from sediment transport in rivers. 3 (2).
Suriñach, E., Flores-Márquez, E. L., Roig-Lafon, P., Furdada, G., & Tapia, M. (2020). Estimation of avalanche development and frontal velocities based on the spectrogram of the seismic signals generated at the Vallée de la Sionne Test Site.(3), 113.
Wenner, M., Walter, F., McArdell, B., & Farinotti, D. (2019). Deciphering debris-flow seismograms at Illgraben, Switzerland.
REFERENCES
Aki, K. and Richards P.G. (2002). Quantitative Seismology (2on Edition). University Science Books.
Brigham, E.O. (1974). The fast Fourier transform. Pearson, 2014
Gauer et al, 2020, https://doi.org/10.1016/j.coldregions.2020.103165
Kanamori and Given., (1982) Analysis of long-period seismic waves excited by the May 1980, eruption of Mount St. Helens- A terrestrial monopole? J. Geophys. Res., 87, 5422-5432
Ichihara, M., Takeo, M., Yokoo, A., Oikawa, J., and Ohminato T. (2012). Monitoring volcanic activity using correlation patterns between infrasound and ground motion, Geophys. Res. Lett., 39, L04304, doi:10.1029/2011GL050542.
Landau, L.D- Lifshift (1987) Fluid mechanics, 2nd ed. Ed. Pergamon press, Oxford
Levy, C.; Mangeney, A.; Bonilla, F.; Hibert, C.; Calder, E.S.; Smith, P.J. (2015) Friction weakening in granular flows deduced from seismic records at the Soufrière Hills Volcano, Montserrat. J. Geophys. Res. Solid Earth, 120, 7536–7557, doi:10.1002/2015JB012151.
Pérez-Guillén, C.; Sovilla, B.; Suriñach, E.; Tapia, M.; Köhler, A.(2016). Deducing avalanche size and flow regimes from seismic measurements. Cold Reg. Sci. Technol. 2016, 121, 25–41, doi:10.1016/j.coldregions.2015.10.004.
Sovilla, B., Köhler, A., Kyburz, M., & Ligneau, C. (2019). The avalanche flow regimes and their pressure on infrastructures. In International Symposium on Mitigative Measures against Snow Avalanches and Other Rapid Gravity Mass Flows (pp. 223-228). Association of Chartered Engineers in Iceland. https://www.dora.lib4ri.ch/wsl/islandora/object/wsl:21560.
Stein, S., Wysession, M., 2003. An Introduction to Seismology, Earthquakes and Earth Structure. Blackwell Publishing.
Suwa, H., Akamatsu, J., Nagai, Y., 2003. Energy radiation by elastic waves from debris flows. Debris Flows Hazard Mitigation: Mechanics, Prediction and Assesement. Millpress, Rotterdam.
Suriñach, E., Tapia, M., Roig, P., and Blach, X. (2018). On the effect of the ground seismic characteristics in the estimation of mass movements based on seismic observation. Geophys. Res. Abstracts, EGU General Assembly 2018,20, https://meetingorganizer.copernicus.org/EGU2018/EGU2018-8479.pdf, https://www.researchgate.net/publication/33128412
Vilajosana, I.; Khazaradze, G.; Surinach, E.; Lied, E.; Kristensen, K. (2007a) Snow avalanche speed determination using seismic methods. Cold Reg. Sci. Technol., 49, 2–10.
Vilajosana, I.; Suriñach, E.; Khazaradze, G.; Gauer, P. (2007 b) Snow avalanche energy estimation from seismic signal analysis. Cold Reg. Sci. Technol., 50, 72–85.
-
AC1: 'Reply on RC1', Emma Surinach, 11 Jan 2023
-
RC2: 'Comment on esurf-2022-10', Anonymous Referee #2, 17 Nov 2022
First of all, I want to kindly thank the authors for their submission of the manuscript entitled “A template to obtain information on gravitational mass movements from the spectrograms of the seismic signals generated” by Suriñach and Flores-Márquez to the journal of Earth Surface Dynamics.
The manuscript discusses the spectral analysis of mass movement signals, and how to extract specific signatures from spectra of seismic data by fitting an exponential to the envelope of the signal onset.
The analysis of the signal onset section is a smart and intuitive way for the classification of seismic events, yet I believe that the manuscript should be shortened significantly and focus on the concise description of the method and/or the given applications before it should be considered for publication. In the sections below, you can find specific comments and a more detailed arguments for my opinion.
One significant issue I see is the similarity between the presented manuscript and the publication by Suriñach et al., 2020 (geosciences). I believe it is important to elaborate on what is actually new in the presented manuscript over the existing publication.
General Comments
- As I understand it, the presented algorithm is fitting an exponential onto the envelope of the signal onset within the spectrogram, resulting in exponents specific for a given type of mass-movements. This clear summary of the algorithm is missing, and instead is often circumscribed with unnecessarily complex paragraphs. Therefore, I believe the manuscript can be shortened significantly by removing redundant sections.
- What is the benefit of using the “template” instead of a thresholding of the fitted exponent values? Is the templating not an unnecessary additional step?
- The introduction reads more like an overview of interesting literature in related fields rather than a focused introduction working towards the main message of the manuscript. I believe that the manuscript would benefit from a more concise and focused introduction.
- If I understand and interpret this correctly, the same method as introduced and used in Suriñach et al., 2020 is used in this manuscript. To me, this was not clear when reading the manuscript. If this is in fact the case, I believe the manuscript could be shortened even more significantly, since the algorithm is already introduced in the above-mentioned publication.
- Please make sure that all acronyms are introduced. For example, PSD (Line 61), STA / LTA (Line 56) are not introduced properly. Also, for algorithms like STA/LTA, it would be great to have a citation in case someone is not familiar with these algorithms.
- The manuscript contains quite a number of redundant statements and paragraphs that contain not much information. Examples are in the paragraphs after line 75 and 125, where “code” and a developed “algorithm” are mentioned, but no information about the content is provided. I believe by removing such redundant and empty paragraphs, the manuscript could be shortened significantly by making it more concise.
- The word “considered” is repeated over and over, even in cases where other words might be more suitable (e.g., data were “used” instead of “considered”, the SON section was “analyzed” instead of “considered").
Specific Comments
Abstract:
- Line 10: you mention a “template” that you simply “overlay” over the data. When only reading the abstract, to me it is quite difficult to grasp what is meant by that. It would be great to get a better intuition of the workflow here, such that the abstract is clearer without reading the entire manuscript.
- Line 14: The way this is formulated is, that if a gravitational mass approaches the sensor, it will start emitting high frequencies. Is this not a path effect that for larger distances, the higher frequencies attenuate faster and do not reach the sensor, hence, the higher frequencies are only measurable closer to the sensor?
- Line 18: The specific values of beta’ in the abstract without context (e.g., an equation, behavior (liner / exponential)) makes it more confusing for me to read. What does this mean? 0.003 / second of what? I would probably explain this more or leave the specific values out.
1 Introduction:
- Line 66: The spectrum is a representation of the time-series data that can be visualized, but it is not really only a “visual representation”?
- Line 75: This paragraph is quite generic and is in my opinion not really needed. Instead, I would briefly mention how you calculate the spectrograms in one sentence.
- Figure 1: Is c) really needed? It contains the same information as in b. Also, in b) the colorbar label is missing.
- Line 125 to 130: Here you redundantly describe that there is a code and an algorithm that was developed. This is rather generic and does not provide any information about the algorithm itself and contains no information.
2 Characteristics of the data used
- Instead of the information collected in text, it would be easier readable if the sensors, dataloggers and characteristics (including references) were in the form of a table.
- 1: This could be re-named to Data pre-processing?
3 Spectrogram vs spectrum
- I believe that this section is not really needed, since the concept of spectra and spectrograms can be referred to as common knowledge in the seismic and seismological community.
- Figure 2: is this Figure really needed?
- Figure 4: Please add colorbar labels. (Please add colorbar labels to all figures and subfigures where they are missing).
4 Spectrogram treatment (ST)
- This is basically the core of the manuscript, describing the algorithm that is used. I believe this chapter could be shortened and better visualized in the form of a simple flowchart.
5 Application
- Figures 7,9 and 11: In the earlier figures, the subfigures were labelled with letters. For consistency, please do this for these figures, too. Colorbar labels are missing.
- I believe that this chapter could also be significantly shortened. Since the manuscripts focus is on the “new method”, I am not convinced that all the experiments and sites need to be introduced in such high detail, but I believe that the summary in Table 6 contains already a lot of the relevant information.
6 Discussion
- I really like table 6 and Figure 15. They are the main take-home messages for me
7 The template
- I have some worries about this section. As I understand, the shape (exponent) of the exponential is dependent on the type of mass-movement, the local geology and morphology (subsurface parameters and incidence angles) and distance of the mass-movement. This makes all the examples in 5. And 6. Site- and event-specific fits of an exponential line to the SON section. I am not yet convinced that the generalization into a template based on these findings adds a substantial benefit. Would it not be better to apply the algorithm in 4. In continuous time-windows to find signal onsets within your data, then calculate the exponents and use these for a classification, rather than “overlying” an arbitrary template on your spectrograms? What is the benefit of this template? I do not really see the reason to do this. Instead of using the template, could not a thresholding of the estimated exponents be a better approach for a classification, since all three investigated events fall within different order of magnitude ranges?
8 Conclusions
- See comments about section 7
- This section does more summarize the entire publication rather than draw conclusions.
- Have you tried the same approach for earthquake recordings or other events ? Might these overlap with the exponents of the shown data ?
- What happens in the absence of an event?
Mentioned literature:
Suriñach, E.; Flores-Márquez, E.L.; Roig-Lafon, P.; Furdada, G.; Tapia, M. Estimation of Avalanche Development and Frontal Velocities Based on the Spectrogram of the Seismic Signals Generated at the Vallée de la Sionne Test Site. Geosciences 2020, 10, 113. https://doi.org/10.3390/geosciences10030113
Citation: https://doi.org/10.5194/esurf-2022-10-RC2 -
AC2: 'Reply on RC2', Emma Surinach, 11 Jan 2023
Dear anonymous reviewer
Thankyou very much for your constructive comments.
Below you will find our replies, but first of all, we want to stress that the template of the envelopes of the spectrograms of the signals we present is a tool for practitioners to be applied in the “classification” to their specific events they are concerned. In principle, avalanche practitioners are not involved in lahars or in other different events and vice versa. Our tool is not for academia, although it is necessary to demonstrate their construction. The objective of this contribution is twofold, one to "academically" demonstrate the validity of obtaining the template, and secondly, to present the template to practitioners.
We have shortened the manuscript as suggested. In the introduction we highlight the difference with our previous paper (Suriñach et al., 2020). Although the method to obtain the spectrogram envelop is the same, the aim of the paper was to study the evolution along the avalanche path of different types of snow avalanches using the envelopes of the spectrograms of the generated seismic signals. We obtained differences in the evolutions of the shape according to the different values of k and beta and we obtained estimation of the speed compared with speeds obtained from external GEODAR measurements. Additionally, we selected the parameter of attenuation coefficient of ground α in the estimation of the avalanche speeds. This is not the aim of the present contribution in which we explain that the increasing shape of the spectrograms is a characteristic independent of the type of mass: snow avalanches at different geographical locations, debris flows and lahars when they approach a seismic sensor. Moreover, we obtain that the values of the 𝜥 and 𝛽 parameters differ between the different types of mass movements (avalanches, lahars and debris flow) and also between them (e.g., avalanche types). Only it is necessary that the mass approaches to the sensor.
As I understand it, the presented algorithm is fitting an exponential onto the envelope of the signal onset within the spectrogram, resulting in exponents specific for a given type of mass-movements. This clear summary of the algorithm is missing, and instead is often circumscribed with unnecessarily complex paragraphs. Therefore, I believe the manuscript can be shortened significantly by removing
We have shortened the manuscript as suggested
What is the benefit of using the “template” instead of a thresholding of the fitted exponent values? Is the templating not an unnecessary additional step?
One of the goals of this contribution is to provide the template as a tool for (technicians) professionals. Once the spectrogram of the generated seismic signal is obtained, professionals can “classify” the event, by using the template. Previously, they must build the template (with different values of 𝛽) and calibrate it.
- The introduction reads more like an overview of interesting literature in related fields rather than a focused introduction working towards the main message of the manuscript. I believe that the manuscript would benefit from a more concise and focused introduction.
Thanks, we have removed some paragraphs in the introduction, and it is centered in the goal of the paper. The goal of our document is to present in a justified manner a template for professionals to apply to the mass movement they are studying or detecting.
If I understand and interpret this correctly, the same method as introduced and used in Suriñach et al., 2020 is used in this manuscript. To me, this was not clear when reading the manuscript. If this is in fact the case, I believe the manuscript could be shortened even more significantly, since the algorithm is already introduced in the abovementioned publication.
In the new version of the introduction, we highlight the difference with our previous paper. Although the method to obtain the spectrograms envelop is the same, the aim of the paper was different. It was to study the evolution along the avalanche path of different types of snow avalanches and in the current study is to present a template that is valid for different gravitational mass movements.
We have eliminated the description of the method and referring to the previous paper.
- Please make sure that all acronyms are introduced. For example, PSD (Line 61), STA / LTA (Line 56) are not introduced properly. Also, for algorithms like STA/LTA, it would be great to have a citation in case someone is not familiar with these algorithms.
PSD ( Power Spectral Density)
STA/LTA Short time average/ long time average (e.g., Allen R.V. 1978) and Vaezi and Vander Baan (2015)
However, we have eliminated this part into the introduction.
- The manuscript contains quite a number of redundant statements and paragraphs that contain not much information. Examples are in the paragraphs after line 75 and 125, where “code” and a developed “algorithm” are mentioned, but no information about the content is provided. I believe by removing such redundant and empty paragraphs, the manuscript could be shortened significantly by making it more concise.
Thanks. We have removed the phrases. However, in the acknowledgments we indicate that the codes were created by our group.
- The word “considered” is repeated over and over, even in cases where other words might be more suitable (e.g., data were “used” instead of “considered”, the SON section was “analyzed” instead of “considered").
Thank you we have replaced these words.
Specific Comments
Abstract:
Line 10: you mention a “template” that you simply “overlay” over the data. When only reading the abstract, to me it is quite difficult to grasp what is meant by that. It would be great to get a better intuition of the workflow here, such that the abstract is clearer without reading the entire manuscript.
- Thank you, Done.
Line 14: The way this is formulated is, that if a gravitational mass approaches the sensor, it will start emitting high frequencies. Is this not a path effect that for larger distances, the higher frequencies attenuate faster and do not reach the sensor, hence, the higher frequencies are only measurable closer to the sensor?
YES. Thank you, we have tried to explain it better
Line 18: The specific values of beta’ in the abstract without context (e.g., an equation, behavior (liner / exponential)) makes it more confusing for me to read. What does this mean? 0.003 / second of what? I would probably explain this more or leave the specific values out.
Ok. Thank you, we have tried to explain it better
1 Introduction:
Line 66: The spectrum is a representation of the time-series data that can be visualized, but it is not really only a “visual representation”?
Yes, you are right, but we use the spectrogram as it is normally represented, as an image. This image contains numerical information, and we are using these values. Note that we mention that the spectrogram is a visual representation of the matrix values of the spectra, and we are using these values.
Line 75: This paragraph is quite generic and is in my opinion not really needed. Instead, I would briefly mention how you calculate the spectrograms in one sentence. Figure 1: Is c) really needed? It contains the same information as in b. Also, in b) the colorbar label is missing.
Figure 1c is obtained with the same information as the spectrogram but represented in 3D waterfall. It was shown to indicate that the spectrogram is not the only manner to visualize the matrix values. However, we have eliminated figure 1c.
We included in the colorbar of all figures (m s-1)2 s. We have changed the dB (without units because it is a relative value) and indicated that the values are in log10 in the figure caption.
Line 125 to 130: Here you redundantly describe that there is a code and an algorithm that was developed. This is rather generic and does not provide any information about the algorithm itself and contains no information.
We have eliminated this paragraph.
2 Characteristics of the data used
Instead of the information collected in text, it would be easier readable if the sensors, dataloggers and characteristics (including references) were in the form of a table.
OK; Table I is incorporated in the text
1: This could be re-named to Data pre-processing?
It is not opportune. In any case it would be pre-analysis of data. Everything mentioned in this section (filtering, FFT, etc.) is data processing like all other processes (ST) before the analysis.
3 Spectrogram vs spectrum
I believe that this section is not really needed, since the concept of spectra and spectrograms can be referred to as common knowledge in the seismic and seismological community.
The seismological community is more familiar with spectrum than with the spectrograms, which are used, normally as the image output of a software package. In our opinion it is better to clarify the differences between them, in order to avoid misunderstandings in the wide audience the journal of Earth Surface Dynamics.
Figure 2: is this Figure really needed?
For us is not needed, because only indicates a change of the reference system but notice the comment of the first reviewer.
We eliminate this figure as suggested.
Figure 4: Please add colorbar labels. (Please add colorbar labels to all figures and subfigures where they are missing).
We included in the colorbar of all figures (m s-1)2 s. We have changed the dB (without units because it is a relative value) and indicated that the values are in log10 in the figure caption.
4 Spectrogram treatment (ST)
This is basically the core of the manuscript, describing the algorithm that is used. I believe this chapter could be shortened and better visualized in the form of a simple flowchart.
We have introduced the following flowchart to clarify the theoretical assumptions and the procedure. In the pdf.
In this new manuscript we have used the term “SON_spectrogram” instead of SON section of the spectrogram.
5 Application
Figures 7,9 and 11: In the earlier figures, the subfigures were labelled with letters. For consistency, please do this for these figures, too. Colorbar labels are missing.
OK done
I believe that this chapter could also be significantly shortened. Since the manuscripts focus is on the “new method”, I am not convinced that all the experiments and sites need to be introduced in such high detail, but I believe that the summary in Table 6 contains already a lot of the relevant information.
The purpose was demonstrated that the behavior of the SON_spectrogram generated from seismic signal generated by mass movements in different situations and sites is the same. This is one of our “findings”. Table 6 without any evidence could be not convincing. We think that for our treatment and values to be credible it is necessary to demonstrate how they have been obtained.
6 Discussion
I really like table 6 and Figure 15. They are the main take-home messages for me
Thank you.
7 The template
I have some worries about this section. As I understand, the shape (exponent) of the exponential is dependent on the type of mass-movement, the local geology and morphology (subsurface parameters and incidence angles) and distance of the mass movement. This makes all the examples in 5. And 6. Site- and event-specific fits of an exponential line to the SON section. I am not yet convinced that the generalization into a template based on these findings adds a substantial benefit. Would it not be better to apply the m in 4. In continuous time-windows to find signal onsets within your data, then calculate the exponents and use these for a classification, rather than “overlying” an arbitrary template on your spectrograms? What is the benefit of this template? I do not really see the reason to do this. Instead of using the template, could not a thresholding of the estimated exponents be a better approach for a classification, since all three investigated events fall within different order of magnitude ranges?
The algorithm 4 is with the amplitudes of the time series not considering frequencies. This is another approach. Note that the exponent corresponds to the increasing shape of the amplitude frequencies in the spectrogram and, not of the amplitudes. When regarding amplitudes intrinsic attenuation is not considered.
One of the purposes of the paper is to supply a template for practitioners. We do not pretend practitioners to adjust numerically their data and found the exponents values. Spectrograms is an output of commercial software easy to obtain. Once the professionals have created and calibrated the template, they can "classify" their event by overlaying and sliding the template over the spectrogram. This allows obtaining the value of beta for a quantitative classification of the type of gravitational mass movement considered.
8 Conclusions
See comments about section 7
This section does more summarize the entire publication rather than draw conclusions.
We have rewritten discussion and conclusion sections
Have you tried the same approach for earthquake recordings or other events? Might these overlap with the exponents of the shown data?
As mentioned in the introduction, the increasing shape of the spectrograms (SON_spectrogram) is due to the existing elastic energy source moving towards a sensor and a product of the attenuation properties of the terrain. This is a peculiarity of mass movements, not necessarily gravitational ones (for example, Almendros et al. 2002). We include a sentence in the discussion.
In the first paragraph of the discussion a comment about this concerns is included.
What happens in the absence of an event?
In this case, no organized signals exist, only noise (wind, anthropogenic …). This is also a characteristic to take into account for the detection. We include a sentence in the discussion.
references
Allen R.V. (1978) Automatic earthquake recognition and timing from single traces Bull. Seism. Soc. Am., 68 1521- 1532
Almendros, J., Ibáñez, J.M., Alguacil, G., and Del Pezzo, E.: Array detection of a moving source, Seismol. Res. Lett. 73, 2 153-165, 2002.
Vaezi, Y and Van der Baan (2015) Comparison of the STA/LTA and power spectral density methods for microseismic event detection . Geophys. Jour. Int. , 203,3,1896-1908. https://doi.org/10.1093/gji/ggv419
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
732 | 201 | 42 | 975 | 37 | 39 |
- HTML: 732
- PDF: 201
- XML: 42
- Total: 975
- BibTeX: 37
- EndNote: 39
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1