the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Sep 2023
Frost quakes in wetlands in northern Finland during extreme winter weather conditions and related hazard to urban infrastructure
Abstract. The paper reports the first results of experiment in northern Finland during winter, 2022–2023, that aimed to study seismic events caused by seasonal freezing in so-called Critical Zone (CZ) of the Earth. These events have attracted public attention recently, as multiple reports about them from local inhabitants in Arctic and sub-Arctic areas appeared recently in social networks. To make an instrumental study of such events, to reveal relationship between their occurrence and winter weather conditions and to evaluate the possible hazard, we installed two high-resolution seismic arrays with co-located soil temperature stations at two sites in Finland, one of them being in the city of Oulu in the sub-Arctic area (65.04° N, 25.61° E) and the other one, above the Polar Circle in municipality of Sodankylä (67.36° N, 26.63° E). The equipment recorded continuous seismic and soil temperature data during November 2022–April 2023. Based on reports from the inhabitants of Talvikangas (Oulu) about ground shaking and unusual noises on 6.1.2023 and their observations of new fractures on the roads there, we selected the time interval for identification of frost quakes originated during that day from continuous seismic records in Talvikangas and in Sodankylä. During the selected time interval, the extremely rapid air temperature drop of about -1.4 °C/hour in Talvikangas and -0.88 °C/hour in Sodankylä were observed. We identified and located two types of seismic events, namely, frost quakes with frequencies of about 10–20 Hz, with waveforms like those of tectonic events, and irregular-shape frost tremors with frequencies of about 120–180 Hz. The sources of frost quakes in Talvikangas are mainly located on irrigated wetland while in Sodankylä about 50 % of registered frost quakes were caused by ice fracturing on the Kitinen river. However, several relatively strong events with origin in wetlands were also recorded. A significant number of sources of frost tremors are confined to wetland areas cut by irrigation channels and to roads cleaned from snow during winter, both in Talvikangas and in Sodankylä. We calculated ground accelerations and ground velocities for strongest events from both groups and compared them to equivalent properties of other seismic signals, like distant local earthquakes in the area, mining production blasts, cargo trains vibration. Our study shows that high-frequency frost tremors corresponding to surface fracture opening in the uppermost frozen surface layer of the thickness up to 5 cm can directly damage infrastructure objects like roads and basements of buildings. Surface waves, produced by frost quakes and propagating inside the shallow soil layer, have large enough ground accelerations at epicentral distances of hundreds of meters. Therefore, frost quakes should be considered as phenomenon, that potentially can damage infrastructures and they have to be taken into account in seismic hazard assessment. Our research is the first instrumental study of seismic events originated from wetland areas. These events occur as a result of interaction between the uppermost layer of the solid Earth’s CZ and atmosphere processes that deserves further study.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(1620 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1620 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-1853', Anonymous Referee #1, 24 Oct 2023
General Comments
Overall, this is a good paper that requires only minor revisions. The paper documents two types of frost-related seismic activity: frost quakes and frost tremors. The methods are adequately described. The evidence from the timing of the events in relation to meteorological data and from the location of the events in relation to water and wetlands make a good case that the signals do indeed derive from the proposed freezing-related mechanisms. The recognition of this phenomenon is significant for the hazard that may it may pose to infrastructure and for understanding the sources of ambient seismic noise in the critical zone. I have only one major comment, regarding the use of HVSR to identify a frequency band of interest (see below), for which the explanation does not really make sense to me.
Specific Comments
Lines 117-124 : I am confused by this. Okkonen et al. (2020) Fig. 2B shows significant vertical as well as horizontal motion, and in the accompanying text they state that the signal is “depleted in body wave energy but having large-amplitude Rayleigh waves”. If the signal is dominated by Rayleigh waves, shouldn’t it have both horizontal and vertical motion and be elliptically polarized in a vertical plane? From your results in this paper, it also looks like it is only the P waves that are horizontally polarized (Figure 6), and the Rayleigh waves as well as the high-frequency tremor in Figure 9 are elliptically polarized. I do agree that the change in HVSR at freezing temperatures is significant, but it could also be explained by a change in the elastic properties of the subsurface due to ground freezing, rather than by a change in the noise source.
Instead of HVSR, or in addition to it, I would suggest plotting a spectrogram of the ambient noise over time. (This could be done with one or both of vertical and horizontal components, or an average of them.) If frost quakes and tremors are a major source of noise in this frequency band, then the same pattern should appear: The noise should occur only when the meteorological conditions are suitable, but the change will be more clearly related to the source, not the medium.
Technical Corrections
- The word “the” is sometimes missing. In the abstract I counted four instances: On, line 8 before “so-called,” line 10 before “relationship,” line 13 before “municipality,” and line 24 before “strongest”. The error doesn't appear to be as frequent in the rest of the manuscript, but it would be good to go through and check.
- Line 26: Missing “and” before “cargo trains”.
- Line 34: Hyphenate "snow-dominated"
- Line 40 “stats” should be “starts”
- Line 80: “is” missing before “located”.
- The Figure 1 caption says “(Mention that units are in metres, X is Easting and Y is Northing)”. This looks like some editing comment that wasn’t fully implemented.
- Line 152: “moth” should be “month”
- Line 167: “such significantly” should probably be “so significantly”.
- Figure 11 caption: The meaning of the blue dot is repeated (and “being” is misspelled in the first instance).
- Line 325: The comma after latitude is not needed.
- Line 362: “shacking” should be “shaking”.
- Line 366: “1-store buildings” should be “1-story buildings”.
- Line 418: “phenomenon” should be “phenomena” since it is referring to two things.
Citation: https://doi.org/10.5194/egusphere-2023-1853-RC1 -
AC1: 'Reply on RC1', Nikita Afonin, 16 Jan 2024
Anonymous Referee #1
Referee: General Comments
Overall, this is a good paper that requires only minor revisions. The paper documents two types of frost-related seismic activity: frost quakes and frost tremors. The methods are adequately described. The evidence from the timing of the events in relation to meteorological data and from the location of the events in relation to water and wetlands make a good case that the signals do indeed derive from the proposed freezing-related mechanisms. The recognition of this phenomenon is significant for the hazard that may pose to infrastructure and for understanding the sources of ambient seismic noise in the critical zone. I have only one major comment, regarding the use of HVSR to identify a frequency band of interest (see below), for which the explanation does not really make sense to me.
Specific Comments
Lines 117-124 : I am confused by this. Okkonen et al. (2020) Fig. 2B shows significant vertical as well as horizontal motion, and in the accompanying text they state that the signal is “depleted in body wave energy but having large-amplitude Rayleigh waves”. If the signal is dominated by Rayleigh waves, shouldn’t it have both horizontal and vertical motion and be elliptically polarized in a vertical plane?
Authors: In the case described in Okkonen et al. (2020) the frostquake was recorded only by single permanent OUL station, located at 14 km from the source. The source of the event appeared as a crack on the road of significant size (about 1m deep and half meter wide). The time of this event was reported by locals as the crack was formed exactly near their house. We found the record of this event on the seismogram of closest seismic OUL station (14 km from source) using this knowledge. On the same day, this station recorded huge number of impulse-like events. They appeared in OUL seismogram during this day only, when the air temperature significantly and abruptly dropped. This was unusual signal, as this station is located in a very quiet area on the bedrock. We analyzed the swarm of these impulses and found out similarity between waveforms of events from the swarm. We noticed from analysis of 3-component recordings that they mainly have horizontal polarization. Assuming that all of these impulses were produced by local sources (because they were presented only by surface waves), we suggested that the main source mechanisms of these events are vertical fracture opening. The recording of the strongest event, mentioned in Okkonen et al. (2020), had strong enough Rayleigh wave. But this is because of the relatively large distance from the source and the effect of propagation. In our pilot study in 2019-2020 as well as in the study described in the current paper, we analyse the events (tremors) with much smaller epicentral distances (less than 1000 m). Assuming that they should have the same polarization as impulses, detected in 2016, we used the HVSR as so-called “polarization filter” to detect frequency band of these events. This is possible because considered tremors look like a continuous swarm composed of events originating almost every second. Therefore, we could consider this process as continuous emission. Calculating the HVSR is working as averaging these impulses over time. Therefore, the HVSR maxima should correspond to the dominating frequency range of impulses.
Referee: From your results in this paper, it also looks like it is only the P waves that are horizontally polarized (Figure 6), and the Rayleigh waves as well as the high-frequency tremor in Figure 9 are elliptically polarized. I do agree that the change in HVSR at freezing temperatures is significant, but it could also be explained by a change in the elastic properties of the subsurface due to ground freezing, rather than by a change in the noise source.
Authors: When we detected such HVSR behavior, the hypothesis about changing mechanical properties of soils was the first we tested. Practically, our first idea was to use the HVSR to understand changing in soils mechanical properties with time (namely, depth of the frozen layer). But when we calculated the time series of HVSR, we found significant disagreements with our original hypothesis. Firstly, the frequency band of HVSR maxima was too low. We had a soil station co-located with the seismic sensor during experiment. From this data we knew that the soil has been frozen only from ground surface down to 5 cm. In that case HVSR peak should be significantly higher (in order of 1kHz). In our case, the frequencies of 120-180 Hz would correspond to a layer of more than 1 m thickness. Moreover, the HVSR maxima in considered frequency range disappeared during about 6 hours after the air temperature became positive. The soil conditions could not be changed so fast. Before making conclusions, we checked the seismograms in the considered frequency range. These seismograms were quite similar to those we observed in 2016: a swarm of horizontally polarized impulses. This significantly supports origin of the HVSR maxima.
Referee: Instead of HVSR, or in addition to it, I would suggest plotting a spectrogram of the ambient noise over time. (This could be done with one or both of vertical and horizontal components, or an average of them.) If frost quakes and tremors are a major source of noise in this frequency band, then the same pattern should appear: The noise should occur only when the meteorological conditions are suitable, but the change will be more clearly related to the source, not the medium.
Authors: We analysed these spectrograms. But we did not detect such obvious correlation between spectrograms and air temperature. This is because of anthropogenic noise, as both sites are seismically not quiet. To find a way to reduce the effect of anthropogenic noise, we used the HVSR method, assuming the source mechanism as vertical opening and taking into account that we record seismograms in near-field area. We found correlation between air temperature and the HVSR behavior. To be sure that the frost tremors correspond to the frequency band detected by the HVSR, we tried to detect tremors in other frequency bands, but without success (they disappeared or no correlation with temperature was observed).
Referee: Technical Corrections
- The word “the” is sometimes missing. In the abstract I counted four instances: On, line 8 before “so-called,” line 10 before “relationship,” line 13 before “municipality,” and line 24 before “strongest”. The error doesn't appear to be as frequent in the rest of the manuscript, but it would be good to go through and check.
Authors: The text has been checked and corrected.
Referee: Line 26: Missing “and” before “cargo trains”.
Authors: Corrected
Referee: Line 34: Hyphenate "snow-dominated"
Authors: Corrected
Referee: Line 40 “stats” should be “starts”
Authors: Corrected
Referee: Line 80: “is” missing before “located”.
Authors: Corrected
Referee: The Figure 1 caption says “(Mention that units are in metres, X is Easting and Y is Northing)”. This looks like some editing comment that wasn’t fully implemented.
Authors: Corrected
Referee: Line 152: “moth” should be “month”
Authors: The typo has been corrected.
Referee: Line 167: “such significantly” should probably be “so significantly”.
Authors: Corrected.
Referee: Figure 11 caption: The meaning of the blue dot is repeated (and “being” is misspelled in the first instance).
Authors: Corrected.
Referee: Line 325: The comma after latitude is not needed.
Authors: Corrected
Referee: Line 362: “shacking” should be “shaking”.
Authors: Corrected
Referee: Line 366: “1-store buildings” should be “1-story buildings”.
Authors: Corrected
Referee: Line 418: “phenomenon” should be “phenomena” since it is referring to two things.
Authors: Corrected
Citation: https://doi.org/10.5194/egusphere-2023-1853-AC1
-
RC2: 'Comment on egusphere-2023-1853', Anonymous Referee #2, 12 Dec 2023
This is an interesting paper on "frost quakes nd tremor" detected in Finland.
I had never heard of such signals before.
The topic is interesting both for seismologists (as a new source of seismic signals), because of the associated damage and hazard and for researchers interested in the Cryosphere.
I recommend this manuscript for publication after minor modification.
I believe that several points should be clarified or improved but that the main results are correct.Earlier experiment during winter 2019-2020 (l115-116).
Could you describe the experimental setting and the results?
l115-120, you wrote that "he frequency band of 120-180 Hz was selected based on results of our earlier pilot experiment in Talvikangas during winter 2019-2020.
In that experiment we recorded three component continuous seismic data by a single station in Talvikangas equipped by broadband Trillium compact seismometer, by Nanometrics Inc. (USA).
The results of our pilot frost quake experiment (Okkonen et al., 2020) suggest that the seismic signal excited by fracturing in the uppermost soils would have mainly horizontal polarization, as the main fracturing mechanism in that case is vertical fracture opening."This paragraph (l115-120) is rather confusing for several reasons.
First, I did not find any description of the 2019-2020 experiment in the paper of Okkonen et al. (2020).
Okkonen et al. (2020) describe data recorded in 2016 with a single sensor. It does not mention tremor.
It also assumes that these quakes were induced by fracture opening based on direct observations of cracks on buildings and roads, not from the analysis of seismic signals.
Is this reference wrong?
Also, the polarisation of the seismic signal depends not only on the source mechanism, but also on the type of waves (P,S, Rayleigh...) and on the source azimuth.
This is explained latter (l195), so that the discussion on the signal polarization could be removed from section 2.
The frequency band of 120-180 Hz mentioned on l115 corresponds to the tremor signal but the quakes have a lower frequency content.
Could you please clarify this part?Figure 2: the largest peak of seismicity occurred on 2016/2/10 but is not discussed in the text.
Do you know what was the source of these events? Are the signals similar to frost quakes?Figure 3b.
Nice tremor signal with a clear correlation with temperature data!
How do you interpret the lower frequency 50 Hz signal when the high-frequency tremor stopped?
Could you show the spectrogram of the signal for the full duration of this 2019-2010 experiment and a zoom on the time window when tremor was detected?
Could you also show a spectrogram for the 2022-2023 experiment for comparison?Source mechanism
Did you try to estimate P wave polarity to constrain the source mechanism?
Vertical motion should be up on all sensors for crack opening. Is it what you observed?
Could you also estimate magnitude (local or moment magnitude)?Temporal evolution of tremor and quakes
Did you observe tremor at both sites?
Fig 13 shows the "number of seismic events": does it includes both quakes and tremors?
Could you show a figure with the number of events per hour for each class (tremor, quakes) and the temperature at each site?Quake detection
l190. What do you mean by "seismograms were ... stacked using beam forming algorithm"?
Could you provide a reference or describe the beamforming method you used?
Is it a simple stack (sum) or do you shift the signals in time? How is this time shift estimated?
Does it really improve the signal quality and the location accuracy compared to picking the raw signals?Figure 6: What are the signal shown in (a): the stack for each array or individual signals? Did you filter the signals?
Same questions for Fig 8.Figure 7. Could you add the sensors on these figures?
Tremor location
The description of the location method is a bit confusing.
Could you explain what signal you correlate? I think you correlate each sensor with the central sensor of the same array but it should be written more explicitly.
The term "virtual source" is also misleading. If I understood correctly this is the central sensor of each array?
Did you also try to use this method to locate frost quakes?Citation: https://doi.org/10.5194/egusphere-2023-1853-RC2 -
AC2: 'Reply on RC2', Nikita Afonin, 16 Jan 2024
Anonymous Referee #2
Referee: This is an interesting paper on "frost quakes and tremor" detected in Finland.
I had never heard of such signals before.
The topic is interesting both for seismologists (as a new source of seismic signals), because of the associated damage and hazard and for researchers interested in the Cryosphere.
I recommend this manuscript for publication after minor modification.
I believe that several points should be clarified or improved but that the main results are correct.Earlier experiment during winter 2019-2020 (l115-116).
Could you describe the experimental setting and the results?Authors: The description has been added to the text.
Referee: l115-120, you wrote that "the frequency band of 120-180 Hz was selected based on results of our earlier pilot experiment in Talvikangas during winter 2019-2020.
In that experiment we recorded three component continuous seismic data by a single station in Talvikangas equipped by broadband Trillium compact seismometer, by Nanometrics Inc. (USA).
The results of our pilot frost quake study (Okkonen et al., 2020) suggest that the seismic signal excited by fracturing in the uppermost soils would have mainly horizontal polarization, as the main fracturing mechanism in that case is vertical fracture opening."This paragraph (l115-120) is rather confusing for several reasons.
First, I did not find any description of the 2019-2020 experiment in the paper of Okkonen et al. (2020).Authors: The paragraph had problems with formulation. From our observations described in Okkonen et. al., 2020, we assumed that the main source mechanism of considered events were vertical opening, as we observed a crack on the road and analysed polarization of impulses in the swarm on 6.01.2016, when the air temperature abruptly dropped and frostquake was reported by locals in Talvikangas. To check this assumption, we organized a pilot experiment in 2019-2020. We installed a broadband seismometer in Talvikangas and detected a swarm of impulses, but in another frequency band (120-180 Hz). They also had horizontal polarization. The swarm was looking like a continuous signal composed of events originating almost every second. That is why we decided to use the HVSR as co-called “polarization filter” to understand behavior of these swarms. Using this approach, we found some correlation with air temperature. Based on this, we assumed that these swarms are frost tremors, but we could not study origin of them, as we had records of only single station in our pilot experiment. When we obtained the data from our experiment in 2022-2023, we checked our hypothesis, by locating sources of signals in considered frequency band. We found that locations of sources correlate with wetlands as well as with reports by local inhabitants about new cracks and noise.
The text has been corrected.
Referee: Okkonen et al. (2020) describe data recorded in 2016 with a single sensor. It does not mention tremor.Authors: The text has been changed. See previous comment.
Referee: It also assumes that these quakes were induced by fracture opening based on direct observations of cracks on buildings and roads, not from the analysis of seismic signals.
Is this reference wrong?Authors: This reference is not wrong, but it is not describing all the details, related to seismic data analysis, as this work aimed to explain mechanism of frostquakes using numerical simulation.
Referee: Also, the polarisation of the seismic signal depends not only on the source mechanism, but also on the type of waves (P,S, Rayleigh...) and on the source azimuth.
This is explained latter (l195), so that the discussion on the signal polarization could be removed from section 2.Authors: That is correct, but in section 2 we explain our assumptions which have been done from direct observations and results of our pilot experiment in 2019 (HVSR). The features of the signals, which we found in 2023 are partially proving our assumptions.
Referee: The frequency band of 120-180 Hz mentioned on l115 corresponds to the tremor signal but the quakes have a lower frequency content.
Could you please clarify this part?Authors: Because the sources of these events are of different scales. In our paper we did not aim to explain everything related to these events. We just detected, located them and estimated ground motions produced by them. We found that only events in frequency bands 2-20 Hz and 120-180 Hz have correlation with air temperature and originating during these extreme weather events.
Referee: Figure 2: the largest peak of seismicity occurred on 2016/2/10 but is not discussed in the text.
Do you know what was the source of these events? Are the signals similar to frost quakes?Authors: They are like impulses from swarm, recorded 6.01.2016. We could not check their origin, because we had records of only single permanent OUL station. We used detector based on correlation analysis to detect them and we just can say that waveforms of these events look quite similar to thoss of events from swarm, recorded in 6.01.2016. However, these events in 2016 motivated us for more detailed instrumental study in 2022-2023, which we report in our present paper.
Referee: Figure 3b.
Nice tremor signal with a clear correlation with temperature data!
How do you interpret the lower frequency 50 Hz signal when the high-frequency tremor stopped?Authors: We investigated the seismograms in the frequency range near 50 Hz as well, but did not find a correlation between air temperature and this maxima behavior. Probably the origin of this maxima is not related to considered phenomena.
Referee: Could you show the spectrogram of the signal for the full duration of this 2019-2010 experiment and a zoom on the time window when tremor was detected?
Could you also show a spectrogram for the 2022-2023 experiment for comparison?Authors: We analysed these spectrograms, but they are not very informative, as both sites are noisy (urban areas). We did not find clear correlation between spectrograms and air temperature. This was one of the reasons why we used the HVSR diagrams instead of spectrograms.
Referee: Source mechanism
Did you try to estimate P wave polarity to constrain the source mechanism?Authors: We did not analyse records in such details. The main goal of current paper is to report about a new type of seismicity we observed as well as about possible hazard, related to this seismicity. We will do detailed analysis of seismograms in our further research, using the whole dataset of our experiment.
Referee: Vertical motion should be up on all sensors for crack opening. Is it what you observed?Authors: We did not analyse records in such details. See previous comment.
Referee: Could you also estimate magnitude (local or moment magnitude)?Authors: In our opinion, ground velocities and accelerations are more informative for seismic hazard assessment in considered cases, as we recorede events in near-field area. In our further study, we are planning to estimate magnitudes as well as to study source mechanisms in more detail.
Referee: Temporal evolution of tremor and quakes
Did you observe tremor at both sites?Authors: Yes (see figure 12)
Referee: Fig 13 shows the "number of seismic events": does it includes both quakes and tremors?Authors: It means number of frost quakes. Fig. 13 caption has been corrected.
Referee: Could you show a figure with the number of events per hour for each class (tremor, quakes) and the temperature at each site?Authors: A sub plot, that illustrates number of frost tremors per hour has been added to figure 13.
In our paper we analysed a single frost quakes episode during 6.01.2023, aiming mainly to locate events and to evaluate ground motion produced by them. As our previous study shows (Okkonen et al., 2020), the air temperature is important factor that could initiate episodes of frost quakes, but it is not the only one. Also snow depth and soil moisture content are important factors. In our next paper, which is now under preparation, we are analysing the whole winter period and the number of events not only for freezing, but also for melting periods. In the next study we take into account not only temperature variations, but also those other factors. The paper will include also detailed thermomechanical modelling. It is not possible to include everything into one single paper.
Referee: Quake detection
l190. What do you mean by "seismograms were ... stacked using beam forming algorithm"?
Could you provide a reference or describe the beamforming method you used?
Is it a simple stack (sum) or do you shift the signals in time? How is this time shift estimated?
Does it really improve the signal quality and the location accuracy compared to picking the raw signals?Authors: We used a method described in Schweitzer et al., 2012 and the reference has been added to the text. We shifted seismograms in time, according to recommendations given there. This procedure helped to improve signal-to-noise ratio and we used these stacked seismograms only on visual inspection (detection) stage. To calculate source location, ground velocities and accelerations, we used seismograms recorded at each sensor.
Referee: Figure 6: What are the signal shown in (a): the stack for each array or individual signals? Did you filter the signals?
Same questions for Fig 8.Authors: Figure 6 shows seismograms, stacked for each array using beamforming, mentioned above. The same is for Figure 8. There are averaged ground velocities and accelerations over the array.
Referee: Figure 7. Could you add the sensors on these figures?
Authors: Done
Referee: Tremor location
The description of the location method is a bit confusing.
Could you explain what signal you correlate? I think you correlate each sensor with the central sensor of the same array but it should be written more explicitly.Authors: That is correct. But this was not always the central sensor. To avoid errors in detection because of low quality of records of central sensor, our algorithm randomly selects several candidates to “virtual source” and tests them. If correlation coefficients are always lower than the threshold (for all candidates to “virtual source”), then the algorithm reports no events in the interval considered. Otherwise, the seismic event is detected and the sensor with the highest correlation coefficient is selected as a “virtual source”.
Referee: The term "virtual source" is also misleading. If I understood correctly this is the central sensor of each array?Authors: This is not always the central sensor, but those with which all other records correlate (the first record in cross-correlation procedure). We are selecting this sensor by tests. If low correlation is detected between all sensors and the central one, the algorithm selects randomly another one as a possible “virtual source” (see previous comment). The term “virtual source” is commonly used in seismic interferometry. We think that usage of this term is justified in our case, as we are using interferometric approach too. We transform seismograms in such a way that after this transformation they look like impulse response of the source, located in point with the “virtual source” cooordinates.
Referee: Did you also try to use this method to locate frost quakes?Authors: We tried, but in our study we preferred to use visual inspection of seismograms and a manual picking of arrivals, as a more reliable method, because we deal with unique data showing quite new phenomenon.
Citation: https://doi.org/10.5194/egusphere-2023-1853-AC2
-
AC2: 'Reply on RC2', Nikita Afonin, 16 Jan 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-1853', Anonymous Referee #1, 24 Oct 2023
General Comments
Overall, this is a good paper that requires only minor revisions. The paper documents two types of frost-related seismic activity: frost quakes and frost tremors. The methods are adequately described. The evidence from the timing of the events in relation to meteorological data and from the location of the events in relation to water and wetlands make a good case that the signals do indeed derive from the proposed freezing-related mechanisms. The recognition of this phenomenon is significant for the hazard that may it may pose to infrastructure and for understanding the sources of ambient seismic noise in the critical zone. I have only one major comment, regarding the use of HVSR to identify a frequency band of interest (see below), for which the explanation does not really make sense to me.
Specific Comments
Lines 117-124 : I am confused by this. Okkonen et al. (2020) Fig. 2B shows significant vertical as well as horizontal motion, and in the accompanying text they state that the signal is “depleted in body wave energy but having large-amplitude Rayleigh waves”. If the signal is dominated by Rayleigh waves, shouldn’t it have both horizontal and vertical motion and be elliptically polarized in a vertical plane? From your results in this paper, it also looks like it is only the P waves that are horizontally polarized (Figure 6), and the Rayleigh waves as well as the high-frequency tremor in Figure 9 are elliptically polarized. I do agree that the change in HVSR at freezing temperatures is significant, but it could also be explained by a change in the elastic properties of the subsurface due to ground freezing, rather than by a change in the noise source.
Instead of HVSR, or in addition to it, I would suggest plotting a spectrogram of the ambient noise over time. (This could be done with one or both of vertical and horizontal components, or an average of them.) If frost quakes and tremors are a major source of noise in this frequency band, then the same pattern should appear: The noise should occur only when the meteorological conditions are suitable, but the change will be more clearly related to the source, not the medium.
Technical Corrections
- The word “the” is sometimes missing. In the abstract I counted four instances: On, line 8 before “so-called,” line 10 before “relationship,” line 13 before “municipality,” and line 24 before “strongest”. The error doesn't appear to be as frequent in the rest of the manuscript, but it would be good to go through and check.
- Line 26: Missing “and” before “cargo trains”.
- Line 34: Hyphenate "snow-dominated"
- Line 40 “stats” should be “starts”
- Line 80: “is” missing before “located”.
- The Figure 1 caption says “(Mention that units are in metres, X is Easting and Y is Northing)”. This looks like some editing comment that wasn’t fully implemented.
- Line 152: “moth” should be “month”
- Line 167: “such significantly” should probably be “so significantly”.
- Figure 11 caption: The meaning of the blue dot is repeated (and “being” is misspelled in the first instance).
- Line 325: The comma after latitude is not needed.
- Line 362: “shacking” should be “shaking”.
- Line 366: “1-store buildings” should be “1-story buildings”.
- Line 418: “phenomenon” should be “phenomena” since it is referring to two things.
Citation: https://doi.org/10.5194/egusphere-2023-1853-RC1 -
AC1: 'Reply on RC1', Nikita Afonin, 16 Jan 2024
Anonymous Referee #1
Referee: General Comments
Overall, this is a good paper that requires only minor revisions. The paper documents two types of frost-related seismic activity: frost quakes and frost tremors. The methods are adequately described. The evidence from the timing of the events in relation to meteorological data and from the location of the events in relation to water and wetlands make a good case that the signals do indeed derive from the proposed freezing-related mechanisms. The recognition of this phenomenon is significant for the hazard that may pose to infrastructure and for understanding the sources of ambient seismic noise in the critical zone. I have only one major comment, regarding the use of HVSR to identify a frequency band of interest (see below), for which the explanation does not really make sense to me.
Specific Comments
Lines 117-124 : I am confused by this. Okkonen et al. (2020) Fig. 2B shows significant vertical as well as horizontal motion, and in the accompanying text they state that the signal is “depleted in body wave energy but having large-amplitude Rayleigh waves”. If the signal is dominated by Rayleigh waves, shouldn’t it have both horizontal and vertical motion and be elliptically polarized in a vertical plane?
Authors: In the case described in Okkonen et al. (2020) the frostquake was recorded only by single permanent OUL station, located at 14 km from the source. The source of the event appeared as a crack on the road of significant size (about 1m deep and half meter wide). The time of this event was reported by locals as the crack was formed exactly near their house. We found the record of this event on the seismogram of closest seismic OUL station (14 km from source) using this knowledge. On the same day, this station recorded huge number of impulse-like events. They appeared in OUL seismogram during this day only, when the air temperature significantly and abruptly dropped. This was unusual signal, as this station is located in a very quiet area on the bedrock. We analyzed the swarm of these impulses and found out similarity between waveforms of events from the swarm. We noticed from analysis of 3-component recordings that they mainly have horizontal polarization. Assuming that all of these impulses were produced by local sources (because they were presented only by surface waves), we suggested that the main source mechanisms of these events are vertical fracture opening. The recording of the strongest event, mentioned in Okkonen et al. (2020), had strong enough Rayleigh wave. But this is because of the relatively large distance from the source and the effect of propagation. In our pilot study in 2019-2020 as well as in the study described in the current paper, we analyse the events (tremors) with much smaller epicentral distances (less than 1000 m). Assuming that they should have the same polarization as impulses, detected in 2016, we used the HVSR as so-called “polarization filter” to detect frequency band of these events. This is possible because considered tremors look like a continuous swarm composed of events originating almost every second. Therefore, we could consider this process as continuous emission. Calculating the HVSR is working as averaging these impulses over time. Therefore, the HVSR maxima should correspond to the dominating frequency range of impulses.
Referee: From your results in this paper, it also looks like it is only the P waves that are horizontally polarized (Figure 6), and the Rayleigh waves as well as the high-frequency tremor in Figure 9 are elliptically polarized. I do agree that the change in HVSR at freezing temperatures is significant, but it could also be explained by a change in the elastic properties of the subsurface due to ground freezing, rather than by a change in the noise source.
Authors: When we detected such HVSR behavior, the hypothesis about changing mechanical properties of soils was the first we tested. Practically, our first idea was to use the HVSR to understand changing in soils mechanical properties with time (namely, depth of the frozen layer). But when we calculated the time series of HVSR, we found significant disagreements with our original hypothesis. Firstly, the frequency band of HVSR maxima was too low. We had a soil station co-located with the seismic sensor during experiment. From this data we knew that the soil has been frozen only from ground surface down to 5 cm. In that case HVSR peak should be significantly higher (in order of 1kHz). In our case, the frequencies of 120-180 Hz would correspond to a layer of more than 1 m thickness. Moreover, the HVSR maxima in considered frequency range disappeared during about 6 hours after the air temperature became positive. The soil conditions could not be changed so fast. Before making conclusions, we checked the seismograms in the considered frequency range. These seismograms were quite similar to those we observed in 2016: a swarm of horizontally polarized impulses. This significantly supports origin of the HVSR maxima.
Referee: Instead of HVSR, or in addition to it, I would suggest plotting a spectrogram of the ambient noise over time. (This could be done with one or both of vertical and horizontal components, or an average of them.) If frost quakes and tremors are a major source of noise in this frequency band, then the same pattern should appear: The noise should occur only when the meteorological conditions are suitable, but the change will be more clearly related to the source, not the medium.
Authors: We analysed these spectrograms. But we did not detect such obvious correlation between spectrograms and air temperature. This is because of anthropogenic noise, as both sites are seismically not quiet. To find a way to reduce the effect of anthropogenic noise, we used the HVSR method, assuming the source mechanism as vertical opening and taking into account that we record seismograms in near-field area. We found correlation between air temperature and the HVSR behavior. To be sure that the frost tremors correspond to the frequency band detected by the HVSR, we tried to detect tremors in other frequency bands, but without success (they disappeared or no correlation with temperature was observed).
Referee: Technical Corrections
- The word “the” is sometimes missing. In the abstract I counted four instances: On, line 8 before “so-called,” line 10 before “relationship,” line 13 before “municipality,” and line 24 before “strongest”. The error doesn't appear to be as frequent in the rest of the manuscript, but it would be good to go through and check.
Authors: The text has been checked and corrected.
Referee: Line 26: Missing “and” before “cargo trains”.
Authors: Corrected
Referee: Line 34: Hyphenate "snow-dominated"
Authors: Corrected
Referee: Line 40 “stats” should be “starts”
Authors: Corrected
Referee: Line 80: “is” missing before “located”.
Authors: Corrected
Referee: The Figure 1 caption says “(Mention that units are in metres, X is Easting and Y is Northing)”. This looks like some editing comment that wasn’t fully implemented.
Authors: Corrected
Referee: Line 152: “moth” should be “month”
Authors: The typo has been corrected.
Referee: Line 167: “such significantly” should probably be “so significantly”.
Authors: Corrected.
Referee: Figure 11 caption: The meaning of the blue dot is repeated (and “being” is misspelled in the first instance).
Authors: Corrected.
Referee: Line 325: The comma after latitude is not needed.
Authors: Corrected
Referee: Line 362: “shacking” should be “shaking”.
Authors: Corrected
Referee: Line 366: “1-store buildings” should be “1-story buildings”.
Authors: Corrected
Referee: Line 418: “phenomenon” should be “phenomena” since it is referring to two things.
Authors: Corrected
Citation: https://doi.org/10.5194/egusphere-2023-1853-AC1
-
RC2: 'Comment on egusphere-2023-1853', Anonymous Referee #2, 12 Dec 2023
This is an interesting paper on "frost quakes nd tremor" detected in Finland.
I had never heard of such signals before.
The topic is interesting both for seismologists (as a new source of seismic signals), because of the associated damage and hazard and for researchers interested in the Cryosphere.
I recommend this manuscript for publication after minor modification.
I believe that several points should be clarified or improved but that the main results are correct.Earlier experiment during winter 2019-2020 (l115-116).
Could you describe the experimental setting and the results?
l115-120, you wrote that "he frequency band of 120-180 Hz was selected based on results of our earlier pilot experiment in Talvikangas during winter 2019-2020.
In that experiment we recorded three component continuous seismic data by a single station in Talvikangas equipped by broadband Trillium compact seismometer, by Nanometrics Inc. (USA).
The results of our pilot frost quake experiment (Okkonen et al., 2020) suggest that the seismic signal excited by fracturing in the uppermost soils would have mainly horizontal polarization, as the main fracturing mechanism in that case is vertical fracture opening."This paragraph (l115-120) is rather confusing for several reasons.
First, I did not find any description of the 2019-2020 experiment in the paper of Okkonen et al. (2020).
Okkonen et al. (2020) describe data recorded in 2016 with a single sensor. It does not mention tremor.
It also assumes that these quakes were induced by fracture opening based on direct observations of cracks on buildings and roads, not from the analysis of seismic signals.
Is this reference wrong?
Also, the polarisation of the seismic signal depends not only on the source mechanism, but also on the type of waves (P,S, Rayleigh...) and on the source azimuth.
This is explained latter (l195), so that the discussion on the signal polarization could be removed from section 2.
The frequency band of 120-180 Hz mentioned on l115 corresponds to the tremor signal but the quakes have a lower frequency content.
Could you please clarify this part?Figure 2: the largest peak of seismicity occurred on 2016/2/10 but is not discussed in the text.
Do you know what was the source of these events? Are the signals similar to frost quakes?Figure 3b.
Nice tremor signal with a clear correlation with temperature data!
How do you interpret the lower frequency 50 Hz signal when the high-frequency tremor stopped?
Could you show the spectrogram of the signal for the full duration of this 2019-2010 experiment and a zoom on the time window when tremor was detected?
Could you also show a spectrogram for the 2022-2023 experiment for comparison?Source mechanism
Did you try to estimate P wave polarity to constrain the source mechanism?
Vertical motion should be up on all sensors for crack opening. Is it what you observed?
Could you also estimate magnitude (local or moment magnitude)?Temporal evolution of tremor and quakes
Did you observe tremor at both sites?
Fig 13 shows the "number of seismic events": does it includes both quakes and tremors?
Could you show a figure with the number of events per hour for each class (tremor, quakes) and the temperature at each site?Quake detection
l190. What do you mean by "seismograms were ... stacked using beam forming algorithm"?
Could you provide a reference or describe the beamforming method you used?
Is it a simple stack (sum) or do you shift the signals in time? How is this time shift estimated?
Does it really improve the signal quality and the location accuracy compared to picking the raw signals?Figure 6: What are the signal shown in (a): the stack for each array or individual signals? Did you filter the signals?
Same questions for Fig 8.Figure 7. Could you add the sensors on these figures?
Tremor location
The description of the location method is a bit confusing.
Could you explain what signal you correlate? I think you correlate each sensor with the central sensor of the same array but it should be written more explicitly.
The term "virtual source" is also misleading. If I understood correctly this is the central sensor of each array?
Did you also try to use this method to locate frost quakes?Citation: https://doi.org/10.5194/egusphere-2023-1853-RC2 -
AC2: 'Reply on RC2', Nikita Afonin, 16 Jan 2024
Anonymous Referee #2
Referee: This is an interesting paper on "frost quakes and tremor" detected in Finland.
I had never heard of such signals before.
The topic is interesting both for seismologists (as a new source of seismic signals), because of the associated damage and hazard and for researchers interested in the Cryosphere.
I recommend this manuscript for publication after minor modification.
I believe that several points should be clarified or improved but that the main results are correct.Earlier experiment during winter 2019-2020 (l115-116).
Could you describe the experimental setting and the results?Authors: The description has been added to the text.
Referee: l115-120, you wrote that "the frequency band of 120-180 Hz was selected based on results of our earlier pilot experiment in Talvikangas during winter 2019-2020.
In that experiment we recorded three component continuous seismic data by a single station in Talvikangas equipped by broadband Trillium compact seismometer, by Nanometrics Inc. (USA).
The results of our pilot frost quake study (Okkonen et al., 2020) suggest that the seismic signal excited by fracturing in the uppermost soils would have mainly horizontal polarization, as the main fracturing mechanism in that case is vertical fracture opening."This paragraph (l115-120) is rather confusing for several reasons.
First, I did not find any description of the 2019-2020 experiment in the paper of Okkonen et al. (2020).Authors: The paragraph had problems with formulation. From our observations described in Okkonen et. al., 2020, we assumed that the main source mechanism of considered events were vertical opening, as we observed a crack on the road and analysed polarization of impulses in the swarm on 6.01.2016, when the air temperature abruptly dropped and frostquake was reported by locals in Talvikangas. To check this assumption, we organized a pilot experiment in 2019-2020. We installed a broadband seismometer in Talvikangas and detected a swarm of impulses, but in another frequency band (120-180 Hz). They also had horizontal polarization. The swarm was looking like a continuous signal composed of events originating almost every second. That is why we decided to use the HVSR as co-called “polarization filter” to understand behavior of these swarms. Using this approach, we found some correlation with air temperature. Based on this, we assumed that these swarms are frost tremors, but we could not study origin of them, as we had records of only single station in our pilot experiment. When we obtained the data from our experiment in 2022-2023, we checked our hypothesis, by locating sources of signals in considered frequency band. We found that locations of sources correlate with wetlands as well as with reports by local inhabitants about new cracks and noise.
The text has been corrected.
Referee: Okkonen et al. (2020) describe data recorded in 2016 with a single sensor. It does not mention tremor.Authors: The text has been changed. See previous comment.
Referee: It also assumes that these quakes were induced by fracture opening based on direct observations of cracks on buildings and roads, not from the analysis of seismic signals.
Is this reference wrong?Authors: This reference is not wrong, but it is not describing all the details, related to seismic data analysis, as this work aimed to explain mechanism of frostquakes using numerical simulation.
Referee: Also, the polarisation of the seismic signal depends not only on the source mechanism, but also on the type of waves (P,S, Rayleigh...) and on the source azimuth.
This is explained latter (l195), so that the discussion on the signal polarization could be removed from section 2.Authors: That is correct, but in section 2 we explain our assumptions which have been done from direct observations and results of our pilot experiment in 2019 (HVSR). The features of the signals, which we found in 2023 are partially proving our assumptions.
Referee: The frequency band of 120-180 Hz mentioned on l115 corresponds to the tremor signal but the quakes have a lower frequency content.
Could you please clarify this part?Authors: Because the sources of these events are of different scales. In our paper we did not aim to explain everything related to these events. We just detected, located them and estimated ground motions produced by them. We found that only events in frequency bands 2-20 Hz and 120-180 Hz have correlation with air temperature and originating during these extreme weather events.
Referee: Figure 2: the largest peak of seismicity occurred on 2016/2/10 but is not discussed in the text.
Do you know what was the source of these events? Are the signals similar to frost quakes?Authors: They are like impulses from swarm, recorded 6.01.2016. We could not check their origin, because we had records of only single permanent OUL station. We used detector based on correlation analysis to detect them and we just can say that waveforms of these events look quite similar to thoss of events from swarm, recorded in 6.01.2016. However, these events in 2016 motivated us for more detailed instrumental study in 2022-2023, which we report in our present paper.
Referee: Figure 3b.
Nice tremor signal with a clear correlation with temperature data!
How do you interpret the lower frequency 50 Hz signal when the high-frequency tremor stopped?Authors: We investigated the seismograms in the frequency range near 50 Hz as well, but did not find a correlation between air temperature and this maxima behavior. Probably the origin of this maxima is not related to considered phenomena.
Referee: Could you show the spectrogram of the signal for the full duration of this 2019-2010 experiment and a zoom on the time window when tremor was detected?
Could you also show a spectrogram for the 2022-2023 experiment for comparison?Authors: We analysed these spectrograms, but they are not very informative, as both sites are noisy (urban areas). We did not find clear correlation between spectrograms and air temperature. This was one of the reasons why we used the HVSR diagrams instead of spectrograms.
Referee: Source mechanism
Did you try to estimate P wave polarity to constrain the source mechanism?Authors: We did not analyse records in such details. The main goal of current paper is to report about a new type of seismicity we observed as well as about possible hazard, related to this seismicity. We will do detailed analysis of seismograms in our further research, using the whole dataset of our experiment.
Referee: Vertical motion should be up on all sensors for crack opening. Is it what you observed?Authors: We did not analyse records in such details. See previous comment.
Referee: Could you also estimate magnitude (local or moment magnitude)?Authors: In our opinion, ground velocities and accelerations are more informative for seismic hazard assessment in considered cases, as we recorede events in near-field area. In our further study, we are planning to estimate magnitudes as well as to study source mechanisms in more detail.
Referee: Temporal evolution of tremor and quakes
Did you observe tremor at both sites?Authors: Yes (see figure 12)
Referee: Fig 13 shows the "number of seismic events": does it includes both quakes and tremors?Authors: It means number of frost quakes. Fig. 13 caption has been corrected.
Referee: Could you show a figure with the number of events per hour for each class (tremor, quakes) and the temperature at each site?Authors: A sub plot, that illustrates number of frost tremors per hour has been added to figure 13.
In our paper we analysed a single frost quakes episode during 6.01.2023, aiming mainly to locate events and to evaluate ground motion produced by them. As our previous study shows (Okkonen et al., 2020), the air temperature is important factor that could initiate episodes of frost quakes, but it is not the only one. Also snow depth and soil moisture content are important factors. In our next paper, which is now under preparation, we are analysing the whole winter period and the number of events not only for freezing, but also for melting periods. In the next study we take into account not only temperature variations, but also those other factors. The paper will include also detailed thermomechanical modelling. It is not possible to include everything into one single paper.
Referee: Quake detection
l190. What do you mean by "seismograms were ... stacked using beam forming algorithm"?
Could you provide a reference or describe the beamforming method you used?
Is it a simple stack (sum) or do you shift the signals in time? How is this time shift estimated?
Does it really improve the signal quality and the location accuracy compared to picking the raw signals?Authors: We used a method described in Schweitzer et al., 2012 and the reference has been added to the text. We shifted seismograms in time, according to recommendations given there. This procedure helped to improve signal-to-noise ratio and we used these stacked seismograms only on visual inspection (detection) stage. To calculate source location, ground velocities and accelerations, we used seismograms recorded at each sensor.
Referee: Figure 6: What are the signal shown in (a): the stack for each array or individual signals? Did you filter the signals?
Same questions for Fig 8.Authors: Figure 6 shows seismograms, stacked for each array using beamforming, mentioned above. The same is for Figure 8. There are averaged ground velocities and accelerations over the array.
Referee: Figure 7. Could you add the sensors on these figures?
Authors: Done
Referee: Tremor location
The description of the location method is a bit confusing.
Could you explain what signal you correlate? I think you correlate each sensor with the central sensor of the same array but it should be written more explicitly.Authors: That is correct. But this was not always the central sensor. To avoid errors in detection because of low quality of records of central sensor, our algorithm randomly selects several candidates to “virtual source” and tests them. If correlation coefficients are always lower than the threshold (for all candidates to “virtual source”), then the algorithm reports no events in the interval considered. Otherwise, the seismic event is detected and the sensor with the highest correlation coefficient is selected as a “virtual source”.
Referee: The term "virtual source" is also misleading. If I understood correctly this is the central sensor of each array?Authors: This is not always the central sensor, but those with which all other records correlate (the first record in cross-correlation procedure). We are selecting this sensor by tests. If low correlation is detected between all sensors and the central one, the algorithm selects randomly another one as a possible “virtual source” (see previous comment). The term “virtual source” is commonly used in seismic interferometry. We think that usage of this term is justified in our case, as we are using interferometric approach too. We transform seismograms in such a way that after this transformation they look like impulse response of the source, located in point with the “virtual source” cooordinates.
Referee: Did you also try to use this method to locate frost quakes?Authors: We tried, but in our study we preferred to use visual inspection of seismograms and a manual picking of arrivals, as a more reliable method, because we deal with unique data showing quite new phenomenon.
Citation: https://doi.org/10.5194/egusphere-2023-1853-AC2
-
AC2: 'Reply on RC2', Nikita Afonin, 16 Jan 2024
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,146 | 213 | 49 | 1,408 | 29 | 30 |
- HTML: 1,146
- PDF: 213
- XML: 49
- Total: 1,408
- BibTeX: 29
- EndNote: 30
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Elena Kozlovskaya
Kari Moisio
Emma-Riikka Kokko
Jarkko Okkonen
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1620 KB) - Metadata XML