Asming V.E., Kremenetskaya E.O.

 

 

Study of applicability of P/S ratio criterion for discrimination of regional earthquakes and explosions in North-Western area.

 

1. Introduction

In this study we consider the possibility of using P/S amplitudes ratio for earthquake-explosion discrimination.

A measured P/S ratio obviously may depend on : kind of the source, its radiation pattern, difference in P and S waves attenuation during their propagation from the source to sensors, difference in P/S ratios caused by local features of the medium near sensors.

We are making an attempt to separate all the factors and study their influence on P/S ratio.

To clearify the possibility to use P/S ratios for discrimination (which seems unclear and doubtful for the authors) we shall try to answer the following questions :

 

1.1. What is the difference between P/S ratios calculated by vertical components and by 3-component stations ?

 

This question is important for seismic arrays typically consisting of sets of vertical sensors. In this case this question sounds as : Is it possible to use vertical sensors of seismic arrays 'as is' or some correction is neccessary ?

 

1.2. How should P/S ratios be measured to achieve the most reliability of the results ?

 

Before making conclusions about physical nature of some events by P/S ratio it is necessery to determine a scheme of measuring P and S amplitudes, i.e., which parts of P and S recordings should be taken into account, is it enough to pick maximal values of amplitudes or it is better to use average values, which frequency ranges are more reliable etc.

 

1.3. What is the variability of P/S ratios measured at different sensors for the same event ?

The following cases are studied :

a) difference in P/S ratios for the same event measured by several vertical sensors (in some cases by several 3-component stations) of the same seismic array (local features of the medium);

b) difference in P/S ratios for the same event measured by different stations (difference in P and S waves attenuation, local geology, probably radiation pattern);

 

1.4. What is the difference of P/S ratios for a group of seismic events of different nature situated rather closely to each others and measured at the same seismic station ?

The purpose is to eliminate the influence of propagation paths, in this case the propagation paths are similar.

 

1.5. Is there at least a statistical difference in P/S ratios for events of different kinds ?

To answer this questions several groups of seismic events are selected (mining explosions, earthquakes, underwater explosions) and estimation of P/S ratios are done for the stations where recordings are available.

2. Method of computations.

 

We made a PC program PScalc for P/S ratios calculation. It enables to input a recording (single channel or 3-component), mark manually 3 time windows : parts of the recordings containing : i) a noice before signal; ii) a part of recording containing P-wave iii) S-wave.

An user can specify a set of band-pass filters to be used for the calculations. After that the signal is passed through each filter of the set. If 3-component recording is in process, its amplitude is calculated as A=Ö(X2+Y2+Z2). To smooth the amplitudes are replaced by STA trace (sliding average of amplitudes by several samples).

Maximal amplitudes for the windows are taken as values for noice, P-wave and S-wave. If the option 'use noice' is on, P and S amplitudes are corrected by noice as Acorrected=Ö(A2-Anoice2). (In fact, this correction appeared negligible for the most part of processed signals).

Optionaly it is possible to use average amplitudes within each time window instead of maximal amplitudes.

These amplitudes are used for P/S ratios calculations. The results can be plotted as P/S ratios or amplitudes versus frequencies (average band frequency is taken) by additional program PSview. (See screenshots in fig.1).

The approach is similar to spectral analysis but gives more smooth results.

In the study we use STA averaging by 10 samples with sampling frequency 40 Hz and the following set of bandpass filters :

F0, Hz

F1, Hz

1

4

1.5

4.5

2

5

2.5

5.5

... etc ...

... etc ...

11.5

14.5

12

15

12.5

15.5

13

16

 

 

 

 

 

Fig.2.1. Screenshots of the programs PScalc and PSview.
3. Variability of P/S ratios measured for the same events at different sensors of the same seismic array.

 

Frode Ringdal noticed [??? reference] that for NORESS recordings P/S ratios of the same seismic event vary significantly from sensor to sensor (vertical sensors). We checked this observation using ARCESS recordings of several events.

The first event we analysed was an underground mining explosion in Khibiny (Kirovsky mine) : 12/10/1997, 67.667 N, 33.727 E, T0=3.37:28, distance to ARCESS is 392 km.

 

Fig. 3.1. Recordings of underground Khibiny explosion 12/10/1997 at several vertical sensors of ARCESS.

 

We processed separately 23 vertical channels and 4 3-c recordings as described in Chapter 2. The results are shown in Figure 3.2.

 

a) Maximal amplitudes of P and S waves for a set of bandpass filters;

b) P/S ratios for the set of bandpass filters and maximal amplitudes;

c) Standard deviations for the two sets of P/S ratios;

 

Fig. 3.2. Results of processing of vertical ARCESS sensors (black lines) and 3-component stations (red lines) for the explosion 12.10.1997

 

The resuls are in a good agreement with the observations made by Frode Ringdal. Indeed, the ratios vary significantly for vertical sensors, less for 3-component stations, although difference in P/S ratios for 3-component stations for some frequency bands amounts to 1.

As it is clearly seen in Fig. 3.2.(a), such variability in P/S ratios are mostly due to P-wave amplitudes variability which is much greater than for S-waves, especially for 3-component stations. (Greater variability of P-waves than S was also noticed by Frode Ringdal in NORESS recordings)

The second event we analyzed was an earthquake under Imandra lake near Khibiny Massif. The earthquake occured 26.06.1996, 21.32:15, coordinates are 67.67 N, 33.00 E, distance to ARCESS about 367 km.

This event is unique : perhaps the only one in North-Western Russia for witch we can completely definitely state that it was an earthquake with a significant depth and a lengthy source, probably displacement along a fault. Our modelling enabled us to estimate its depth as 15-20 km (see Fig 3.3). We shall refer to the earthquake in other chapters.

Fig 3.3. Recordings of the 26.06.1996 earthquake by vertical sensors of Apatity array (epicentral distance about 16 km). A strong difference in recordings by closely located sensors indicates a lengthy source.

 

18 vertical sensors and 2 3-component stations (ARCESS) have been processed. The results are shown in Fig.3.4.

a) Maximal amplitudes for set of bandpass filters;

b) P/S ratios by maximal amplitudes;

Fig. 3.4. Results of processing of vertical ARCESS sensors (black lines) and 3-component stations (red lines) for the Imandra earthquake 26.06.1996

 

The results are similar : great difference in P/S ratios for vertical sensors, less but significant for 3-component stations, difference in ampitudes, especially for P-wave.

One can of course try to explain this strange phenomenon by deviations in amplitude-frequency responses of different sensors. But this is doubtful because the amplitude curves have different shapes and this does not explain the difference in P/S ratios.

Our explanation is that such difference is due to the conductive medium properties in the vicinity of the sensors (geological features and/or relief). To check this we calculated average P/S ratios by all frequency bands for each sensor and plotted them as points of different colors in coordinates of the sensors. The results are shown in Fig. 3.5.

Fig. 3.5. Average P/S ratios by vertical sensors of ARCESS for Imandra earthquake 26.06.1996 (left) and Khibiny underground explosion 12.10.1997 (right). Red points mark sensors with higher P/S ratio. Lines indicate azimuth to the events.

 

Some regularity does exist, indeed, blue circles indicating low P/S ratios are grouped mostly on the right. For our mind it confirms that the difference in P/S ratios is mostly due to the local inhomogeneity of the medium.

We have no exact explanation of the phenomenon that scattering of S-wave amplitudes is much less than for P-waves.

Our assumption is as follows : if a wave meets a surface near a sensor it can produce a secondary wave, say P may produce S and vice versa. This waves move with different speeds and therefore arrive to the sensor with a time delay. The seismic energy carrying by a wave is dividing into two groups separated in time so when we are calculating maximal amplitudes we may take into account only one of these groups.

If the producing of secondary waves is not symmetrical (S waves produce less P than P produce S) it can explain the situation.

No matter is this explanation close to the truth or not it led us to the conclusion that it is preferably not to calculate maxima of amplitudes within time windows but use averaged values for the windows. Unlike the maxima this approach require accurate definition of the time windows. We used for the two events time windows for P covering the most part of Pn starting with its onset and corresponding windows for S about 1.7 times longer (Vp/Vs).

The results (for all 3-component stations) including ratios and amplitudes calculated by maxima, for vertical sensors and whole stations as well as amplitudes and ratios calculated by the averaging are shown in Figures 3.6 and 3.7.

 

Fig. 3.6. Underground explosion in Khibiny 12.10.1997. Comparison of P and S amplitudes calculated by :

- maxima, vertical sensors (blue thin lines);

- maxima, 3-component (red thin lines);

- averaging, 3-component (black thick lines);

a) amplitudes;

b) P/S ratios;

c) Standard deviations for P/S ratios

 

 

Fig. 3.7. Imandra earthquake 26.06.1996. Comparison of P and S amplitudes calculated by :

- maxima, vertical sensors (blue thin lines);

- maxima, 3-component (red thin lines);

- averaging, 3-component (black thick lines);

 

The scattering in P amplitudes as well as in P/S ratios diminished significantly in both cases. Maximal difference in P/S ratios is about 0.3 in first case and 0.1 in second case.

But we still did not checked the situation where the averaging is applying to Z-components only. The figure below presents the comparison of using the averaging for 3-component stations and vertical sensors only :

 

Fig. 3.8. P/S ratios and amplitudes by averaging using vertical sensors (red lines) and 3 components (black thick line). Khibiny explosion 12.10.1997

 

It is clearly seen that for the approach with averaging scattering of amplitudes is essentially greater for vertical sensors only than for 3 components.

 

Finishing the chapter, we compare P/S ratios for the two considered events using the most stable method namely average amplitudes by 3-component stations.

Fig 3.9. P/S ratios calculated by 3-component stations (all available 3-c data were used) for the underground explosion 12.10.1997 (black lines) and Imandra earthquake 26.06.1996 (blue lines).

 

As it is seen these two events may be discriminated well in high enough frequency bands. But note, that these two events have very similar event-to-station paths and very different nature.

 

Conclusions

 

- single vertical sensors are completely useless for P/S ratio estimating;

- approach with calculating average amplitudes in compatible time windows (under compatible we mean starting with wave onsets and with length of S-window about 1.7 times greater than P), in contrast to the calculation of maxima, enables to diminish significantly the influence of local inhomogenity on the P/S ratio estimate;

- the variability of this approach requires some more detailed statistical study;

 

 

4. Variability of P/S ratios in some special cases.

 

In previous chapter we considered P/S ratios for recordings of the same seismic event at the same station (different sensors of a seismic array) and worked out an approach of P/S estimation which seems to be the most reliable to diminish the influence of local inhomogeneity of the medium.

In this chapter we shall apply the approach to several special cases to highlight the influence of another factors mentioned in the introduction. In general, the following cases should be considered :

 

- one KIND of events occured in the same place - one stations (variations in P/S ratios within a discrimination group). Mining explosions in Khibiny recorded at ARCESS will be considered;

 

- one event - different stations (influence of propagation paths);

 

- one KIND of events occured in different places - one station (Mining explosions in Kola Peninsula recorded at ARCESS will be considered);

 

- one KIND of events - different stations;

 

- different kinds of events - one station;

 

- different kinds of events - different stations (to check how the criterion could be applied as is, i.e., without adjusting calibrations for the mentioned factors;

 

Not all of the cases mentioned will be considered in the study. Some of them will be the matter of future investigation.

 

4. Sets of events occured closely to each other and recorded by the same stations.

 

We selected 7 Khibiny mining explosions (both open-pit and underground) well-recorded at ARCESS.

We added the recording of mining collapse occured 16.04.1989 immediately after an underground explosion in Kirovsky mine.

And 3 more events are strongly suspected to be rockbursts in the Khibiny Massif within the mining area.

For comparison the earthquake 26.06.1996 (path to ARCESS is near the same) was added.

The events are given in the table below (distances to ARCESS are 380-400 km).

 

T0

Mine

Remark

28.07.1996 03.14:30

Yukspor

Underground, 95 ton

31.07.1996 11.09:25

Koashva

Open-pit, 179 ton

02.08.1996 11.19:35

Central

Open-pit, 113 ton

02.08.1996 11:23:52

Central

Open-pit

16.08.1996 08.23:55

Koashva

Open-pit, 94 ton

29.09.1996 06.05:48

Kirovsky

Underground, 350 ton

12.10.1997 03.37:30

Kirovsky

Underground

 

 

 

26.10.1995 05.31:06

Rasvumchorr-Central

Rockburst

17.11.1993 07.45:14

Kirovsky

Rockburst

12.12.1993 08.50:05

Kirovsky-Yukspor

Rockburst

 

 

 

26.06.1996 21.32:15

67.67 N, 33.00 E

Imandra earthquake

distance to ARC = 367 km

 

 

16.04.1989 06.34:44

Kirovsky

Mining collapse

 

We computed P/S ratios by the central 3-component station of ARCESS. The results are given in Figure 4.1.1. Note, that P/S ratios for the mining collapse, the rockbursts and the earthquake are not included into standard deviation (i.e., it is computed by explosions only).

Fig. 4.1.1.

P/S ratios for Khibiny mining explosions (balck lines), the mining collapse 1989 (magenta line), mining rockbursts (red lines) and Imandra earthquake (blue line) on top.

Standard deviation of P/S ratios for the explosions;

 

It is easy to make a preliminary observation : the most part of explosion curves (excluding the first explosion and the collapse) coincide well in low and high frequency bands, having a significant scattering from 6 to 10 Hz. All explosions curves coincide well in high frequencies.

The curves for the mining collapse as well as for the mining rockbursts are situated below the curves for the explosions in higher frequency bands. The curve for Imandra earthquake is even lower.

 

The second group of events comprises 3 mining explosions in Zapolyarny (about 200 km from ARCESS) :

23.08.1996 11.11:30

23.08.1996 11.13:56

28.08.1996 11.23:37

 

Fig 4.1.2. P/S ratios for 3 Zapolyarny explosions recorded by ARCESS and their standard deviation.

 

Although all 3 curves are similar there is a great difference if to compare with the curves for Khibiny events. How to explain the difference (by difference in the sources character or by different path-to-station influence or by some combination of the factors) is the matter for future study.

 

4.2. Recordings of events of similar types (but occured in different places) for the same station.

 

Mining explosions in Kola Peninsula and Kareliya recorded at ARCESS are considered. In addition to the Khibiny mining explosions mentioned above the explosions in Kostomuksha, Olenegorsk and Zapolyarny (the 3 explosions mentioned above) has been processed.

 

Extra explosions :

DATE

TIME

COORD

DIST TO ARC

REMARK

22.08.1996

08.24:42

68.048 N 33.145 E

349 km

Olenegorsk

28.08.1996

09.49:02

64.697 N 30.689 E

582 km

Kostomuksha

 

 

The figure with the results (excluding two mostly deviant Khibiny explosions curves) is given below. For comparison Imandra earthquake 26.06.1996 was added.

 

Fig. 4.2.1. Mining explosions in Kola Peninsula recorded by ARCESS plus Imandra earhquake (top) and standard deviation of P/S ratios for all the explosions (bottom).

 

The significant difference between the groups of curves are seen. Is it due to the influence of propagation paths or difference in physical nature of the explosions (in the exploding technology) should be studied in future.

 

 

 

 

4.3. Recordings of the same event at different stations.

 

Below we present results of processing of several seismic events, both earthquakes and explosions, each one by all stations which data we have.

The first is event of unknown nature occured 16.08.1997 in Kara Sea near Novaya Zemlya : 16.08.1997 2.11:00, 72.47 N 57.76 E.

 

 

Fig 4.3.1. Event 16.08.1997 in Kara Sea

 

Unfortunately data by NRI and KEV had digitizing frequency 20 Hz and were computed in half of the frequency bands. Stations KBS and SPI are closely to each other. Nevertheless, for far stations KEV, KBS and SPI the curves are closely to each other both in lower and higher frequency bands (just the same was observed for Khibiny explosions) and differ in the middle.

Significant difference between these curves and the Amderma curve probably demonstrates a dependence of P/S ratio from distance.

 

Next we considered the mentioned above calibration underground explosion in Kirovsky mine in Khibiny (29.09.1996 06.05:48). It was recorded by many seismic stations, we selected the recordings with high signal-to-noise ratio, including recordings at temporary station PLQ which was installed specially to register the explosion.

Fig 4.3.2. Calibration underground explosion in Khibiny

 

In this case results are variable and do not show any clear dependence of P/S ratio on distance.

 

An extremely interesting event for our study is the earthquake occured 21.01.1996, 2.16:34 in Northern Norway (69.156 N, 24.516 E). By some estimates its depth was about 10-20 km. The results of its processing are shown in Figure 4.3.3

Fig 4.3.3. Earthquake in Northern Norway, 21.01.1996.

 

The results can be easy divided onto two groups of curves with very similar shapes, each one containing both far and near stations : 1) AMD (1425 km) and KEV (118 km) and 2) KBS (1144 km), LVZ (435 km) and AP0 (387 km).

The only explanation we can give is that it is due to non-symmetrical source, i.e., different radiation patterns. The geometry of the source and stations is shown in Figure 4.3.4.

Fig. 4.3.4. Location of the 21.01.1996 earthquake and directions to the stations.

 

Indeed, athimuths to stations KEV and AMD are close, same as to LVZ and AP0 and azimuth to KBS is almost opposite to AP0.

 

The situation is very similar for the Imandra earthquake mentioned several times above (26.06.1996, 21.32:15, 67.67 N, 33.00 E). The results for stations LVZ and AP0 (curves of the same shape) and ARC, as well as the map of the event and azimuth are given in Figure 4.3.5.

 

 

Fig. 4.3.5. P/S ratios for Imandra earthquake 26.06.1996 and configuration of the event and stations;


DISCUSSION AND CONCLUSIONS

 

1. P/S ratios calculated by vertical sensors recordings and/or maximal P and S amplitudes are extremely variable due to the medium properties in the vicinity of sensors. Such variability can be significantly diminished by using 3-component recordings and average P and S amplitudes for compatible time windows.

 

2. Frequency patterns of P/S ratios of events of the same type recorded by the same station seem to be close to each other (example of Khibiny explosions recorded at ARCESS) whereas seismic events of other types (the mining collapse, rockbursts and Imandra earthquake) occured about in the same place differ from explosions, especially in higher frequency bands.

The observed difference was regular : Khibiny explosions have higher P/S ratios, less for rockbursts and even more less for the Imandra earhquake.

 

3. P/S ratios for events of similar types occured in differnt places (examle of mining explosions at Kola Peninsula) and recorded by different stations have a great scattering. Possible reasons are : influence of propagation paths; different exploding technologies. This requires additional study.

 

4. The great scattering of P/S ratios was observed for the case one event - many stations. For 2 earthquakes the scattering can be clearly separated onto two parts : a) due to non-symmetrical radiation patterns of their sources; b) dependence on distance.

 

The following preliminary conclusions can be done :

 

1. It seems quite impossible to use P/S ratio for events discrimination as is, i.e., by recordings of a single station without preliminary calibration.

 

2. Even after preliminary calibration the possibility of use P/S ratio is doubtful for a single station because some non-symmetrical sources can be wrongly discriminated due to their radiation patterns influence.

 

3. It may appear to be possible to use P/S ratios for some predefined set of seismic stations for the discrimination of events within a predefined region after carefull calibration, i.e., taking into account all the factors mentioned above

 


APPENDIX

 

Seismic stations and seismic events used in the study

 

Table A1. Seismic stations

Name

Coordinates

Remark

ARC

69.5349 N 25.5058 E

ARCESS, array

AP0

67.6033 N 32.9944 E

Apatity, array

AMD

69.7440 N 61.6480 E

Amderma, micro-array

LVZ

67.8979 N 34.6514 E

Lovozero, 3-C

NRI

69.4000 N 88.1000 E

Norilsk, 3-C

KBS

78.9180 N 11.9240 E

Spitsbergen, 3-C

SPI

78.1777 N 16.3700 E

Spitsbergen, array

KEV

69.7550 N 27.0070 E

3-C

PLQ

66.4100 N 32.7500 E

Temporary 3-C station

 

Fig. A1. Map of seismic stations used in the study.

 

 

 

Table A2. Seismic events

Date

T0

Coordinates

Remark

16.04.1989

06.34:44

 

Mining collapse, Kirovsky

17.11.1993

07.45:14

 

Kirovsky, rockburst

12.12.1993

08.50:05

 

Kirovsky-Yukspor,

rockburst

26.10.1995

05.31:06

 

Rasvumchorr-Central,

rockburst

21.01.1996

2.16:34

69.156 N, 24.516 E

Northern Norway

26.06.1996

21.32:15

67.67 N, 33.00 E

Imandra earthquake

distance to ARC = 367 km

22.08.1996

08.24:42

68.048 N 33.145 E

Olenegorsk

23.08.1996

11.11:30

 

Zapolyarny

23.08.1996

11.13:56

 

Zapolyarny

28.08.1996

09.49:02

64.697 N 30.689 E

Kostomuksha

28.08.1996

11.23:37

 

Zapolyarny

16.08.1997

2.11:00

72.47 N 57.76 E

Kara Sea

12.10.1997

03.37:28

67.667 N, 33.727 E

Underground calibration Khibiny Kirovsky