Asymmetric Impulses

Figure 5
[35]  Method 1 was used to separate possible high-amplitude impulses in the minute range of periods. In Figure 5 the results are given of processing initial 20 Hz records of five stations (Figure 1) before Kronotskoe earthquake in which components were separated with periods ranging from 8 to 128 minutes. ARU data were not used because of the noise of technogenic origin in the day time. From the figure, it follows that in the time interval under investigation 84 hours before the earthquake no considerable variation of noise level was noted in any of the five stations. The amplitude of impulses at each of the stations only makes several percent of the level of common microseisms in the range of 2-6 seconds. We studied the variation of noise level with the period increase in all of the stations. It was revealed that microseisms have maximum amplitude in the range of 3-5 s, which is in agreement with the model of their oceanic origin. As the period lengthens the amplitude drops, so that with period of 1 minute it, on the average, decreases by 20 times. Then the decrease of amplitude slows down and after a period of 5-10 minutes it gradually starts growing. In the qualitative sense, the spectra of oscillations intensity have the same form as the spectrum for PET station that is shown in Figure 4. However, the farther is the station from PET, the slower is the decrease with lengthening of periods in the range from seconds to first minutes. Thus values of oscillations velocities shown in y -axis of plots in Figure 5 allow us to draw the following preliminary conclusion. The fact that their value gradually decreases as the distance of the station from Kamchatka (PET) increases and their level is higher for an order of magnitude in PET station as compared to other stations suggests that their source was located in the Pacific seismically active zone.

Figure 6
[36]  The feature of PET station is asymmetric impulses of negative polarity. The dynamics of the impulses quantity change automatically separated by the program in 30-day interval before Kronotskoe earthquake is shown in Figure 6. Curve 1 shows the level of "common'' microseisms of second range, where D is dispersion of initial record in successive windows of length of 4 seconds (80 initial counts with sampling frequency of 20 Hz). Two upper curves show variations of the amount of impulses found by the program and having positive and negative polarity of minute range for 1 day with a step of 0.1 day. Gradual increase in the amount of impulses is noted especially of negative polarity N(-) before the earthquake.

[37]  We compared oscillations of the amount of asymmetric impulses and the level of microseisms of second range of periods; no correlation between them was established. It means that impulses were not caused by storms in the seas and oceanic areas. The influence of more distant meteorological phenomena, for example in the Atlantic regions, is not excluded, but this possibility was not checked. Correlations between the quantity of impulses and semidiurnal, diurnal and two-week earth tides was not revealed either.

Figure 7
[38]  Let us consider the structure of individual impulses and their sequence. One and the same section of PET station record of 3 December of about 3-hour duration is shown in plots 1 and 2 (Figure 7). The lower plot is the record with discreteness 1 s filled with microseism variations of the second range of periods. Plot 2 is the low-frequency constituent after high frequencies were rejected with smoothing by Gaussian kernels with averaging radius of 100 s. In this plot you can see impulses with the following features: 1) duration of individual impulse is ~10-15 min, and intervals between them make ~40 min; 2) The impulse form changes from almost symmetrical to one-polar with gradual suppression of positive phase. Note that impulse amplitude in plot 2 is an order less than microseisms level in plot 1 and therefore they cannot be seen in the lower plot. A record fragment of the same time and processed in a similar way from Obninsk station located at the East European platform at a distance of 6500 km from Petropavlovsk station (see Figure 1) is shown in plot 3 (Figure 6) for comparison. Impulses of the type of plot 2 are not detected and the amplitude of oscillations is two less orders.

Figure 8
[39]  Similar results were obtained before Neftegorskoe earthquake for Yuzhno-Sakhalinsk station (YSS), which was the nearest to the epicenter. In Figure 8, record fragments of microseisms (plot 1) and of low-frequency component (plot 2) of this station and of Obninsk station (plot 3) for comparative purposes are shown. In plot 2 both asymmetric and symmetric impulses of duration ~10-15 minutes are separated. Comparing Figure 8 and Figure 7 we note the following: durations of individual impulses (plots 2) coincide in both cases; as distinct from Kronotskoe earthquake, intervals between sequential impulses before Neftegorskoe earthquake are not regular; impulse amplitude in Figure 8 is approximately one twentieth as much as compared to Figure 7; microseisms amplitude (plot 1 Figure 8) is approximately one twenty-fifth as much as the one in Figure 7; oscillations amplitude at OBN station (plots 3) is comparable in the both cases (~5-7 nM s -1 ), and their structure differs from that of stations YSS and PET. Before Neftegorskoe earthquake the amount of negative polarity impulses increased in the last 5 days against the undisturbed background of second-range microseisms.

Figure 9
[40]  Before Hokkaido earthquake the records of two stations shown in Figure 2 intense impulses were revealed with amplitudes above diurnal and semidiurnal tidal oscillations. It is demonstrated in Figure 9, where 4-day records (96 hours) are given for the period of 16-19 September 2003: plot ERM is of Erimo station located in the southeast of Hokkaido in subduction zone and actually in the epicentral zone of the earthquake; plot PET is of station Petropavlovsk on Kamchatka shore located in subduction zone; plot MDJ is of station Mudanjiang located in the continent in northeastern China. In ERM and PET plots, individual impulses of positive and negative polarity as well as series of impulses close in time can be seen against diurnal and semidiurnal tides. Positive polarity series is noted in the area of 16-20 hours in ERM plot; negative polarity series in PET plot takes the interval of 82-96 hours. We note three features that in our opinion are significant: 1) impulses were only manifested at stations ERM and PET located in subduction zone, 2) the times of both single impulses and series of impulses do not coincide in ERM and PET plots, 3) impulses amplitude in ERM plot is on the average greater as in magnitude and with respect to the span of tidal variations both as compared to plot PET. From data given in Figure 9, it is reasonable to assume that the sources of impulse oscillations are located in subduction zone. Since we do not know the actual sensitivity of a number of stations in the minute and hour range of periods, we give attention to relative amplitudes of oscillations in comparison with tides or microseisms of second range.

Figure 10
[41]  More detailed structure of impulses of stations ERM and PET is shown in Figure 10. We changed impulses polarity in ERM plot artificially to inverse as compared to plot 9 so that comparing is more convenient. Low-frequency oscillations of hour range of periods caused by earth tides were rejected by deducting Gaussian trend with the radius of 1000 s. Intervals between sequential impulses make first thousands of seconds. From more detailed time base of the impulses 1, 2, 3, and 4, it can be seen that the time of impulse rise to extreme values is not the same and lies in the interval from 100 to 200 seconds. This interval is within standard frequency range of IRIS stations IRIS [Starovoit and Mishatkin, 2001]. From Figure 10 it follows that the comparative amplitude of impulses with respect to high-frequency noise of second-range microseisms is greater at ERM stations. Since BHZ measuring channels of IRIS stations record the vertical component of displacement velocity, it is believed that the presented impulses correspond to one-polar vertical step of the base movement. Comparison with horizontal component record showed that horizontal component amplitudes of N-S and E-W impulses are almost one less order as compared to vertical components and are comparable to high-frequency noise amplitudes.

[42]  Before Sumatra earthquake, quasi-sinusoidal symmetric oscillations in the minute range of periods were only detected in the records of stations shown in Figure 3.

Figure 11

Periodic Oscillations

[43]  As a result of applying method 2 to the analysis of microseism variations before Kronotskoe earthquake of 5 December 1997 with M = 7.8 and coordinates [54.64oN, 162.55oE], spectrum-time diagrams of logarithmic likelihood function increment D were obtained for six seismic stations IRIS: PET, YAK, OBN, MAG, YSS, and ARU that have similar characteristics and the location of which is shown in Figure 1. The stations are located in different seismic geological settings at considerable distances from each other. Station PET is located in subduction zone at a distance of D = 310 km from the epicenter. Station MAG second closest to the epicenter ( D = 900 km) is located in the north of the Sea of Okhotsk characterized by deep earthquakes. In the records of those two stations, periodic oscillations were revealed three hours before Kronotskoe earthquake (Figure 11). Red bands on the figure indicate emergence of periodic oscillations in the period range from 20 to 60 minutes. Increment D was estimated for a sequence of time moments of seismogram maximums exceeding the level equal to the mean value in the window plus sample standard deviation in the same window. The count starts from 00 Greenwich time (UTC) 2 December 1997. Values D were calculated in time window of duration DT = 3 hours with a shift DS = 1 hour, thus diagrams are presented in the figure, beginning at 3 o'clock 2 December. Last points in the time scale in diagrams (83 hours) correspond to 11 hours on 5 December (27 minutes before the moment of earthquake).

[44]  The beginning of the period under investigation was chosen because 2 December 1997 was characterized by quiescent seismic conditions were noted in the zones where the enumerated-above stations are located and where local earthquakes of energy class K>10 ( M>4 ) were not noted from Geophysical Service RAS data. Dramatic activation of the seismic process in the future Kronotskoe earthquake epicentral area started in Kamchatka in the middle of 3 December. That day three earthquakes of K>10 and three earthquakes of K>11 occurred there. The beginning of foreshock process is marked with an arrow and symbol F in the upper diagram of Figure 11. The activation went on with increasing number of shocks on the following day; seven events with K>10, five events with K>11 and one with K=12.8 ( M = 5.5) were recorded on 4 December. The latter is indicated with an arrow and symbol F a in Figure 5. On the day of Kronotskoe earthquake, 13 foreshocks K>10, 12 foreshocks of K>11 and 4 foreshocks of K>12 including one of K=12.5 ( M = 5.3) were registered on 5 December before the earthquake moment. The latter is indicated with an arrow and symbol F b in Figure 11. Altogether the foreshock series included 100 earthquakes of representative class K>8.5 ( M>3 ).

[45]  From Figure 11 it follows that the first series of periodic microseismic oscillations manifested itself at the end of 2 December - beginning 3 December. The second series started approximately at 2200 on 4 December (71 hours in diagrams) and was in progress until the time of the Kronotskoe earthquake main shock. Comparison was made with data of stations YAK, YSS, ARU, OBN, which are more remote from the source [Sobolev et al., 2005]. It was revealed that the first series of periodic oscillations was appeared in diagrams of D of 2-3 December. In the second series, oscillations at those stations were noted immediately after foreshock F a, but they were missing after foreshock F b up to the earthquake moment. Since periodic oscillations 3 hours before the earthquake were only revealed at stations PET and MAG, which are the nearest to the epicenter, it is reasonable to assume their relation to the process in the subduction zone adjacent to the epicenter. The effect was most pronounced in PET station and maximum D indicated a period of ~37 minutes 1 hour before the earthquake.

[46]  Emergence of periodic oscillations in minute range of microseisms was revealed before Sumatra earthquake of 26 December 2004 with M = 9.2 and epicenter coordinates [3.32oN, 95.85oE]. Preliminary analysis of records made by stations in Figure 3 showed that station MBWA in Australia had not been operated in the period of Sumatra earthquake, there had been failures and gaps in the records of DGAR and PALK stations and stations DAV and QIZ located in the Pacific region had a different structure of microseismic oscillations as compared to stations located in the Indian Ocean region. Therefore major studies were carried out from the data of stations CHTO, KMI, XAN, COCO, and partially PALK. We used records of vertical components with the exception of COCO station where this component had not been registered; in the latter case, horizontal component data were processed. The base of data of those stations that we used covered the interval from 6 to 26 December (00 hours 58 minutes) of that is to say until the moment of Sumatra earthquake.

[47]  A singular feature was the fact that 2.5 days before Sumatra earthquake in the southern hemisphere another strong earthquake had occurred of M = 7.9; the epicenter of the earthquake with coordinates [49.31oS, 161.35oE] was located to the southwest of New Zealand (in Makkuori ridge area). Oscillations of this earthquake were hundreds of times as much as microseisms level in the above-mentioned stations.

Figure 12
[48]  Time-frequency diagrams D shown in Figure 12 for stations KMI and CHTO were calculated with the use of method 2 described above. Arrows indicate the time when Sumatra earthquake ( M = 9.2) and preceding Makkuori earthquake ( M = 7.9) occurred. Periodic oscillations appeared after Makkuori and continued for one a day. Comparison with Figure 11 suggests an effect similar to the one noted after the foreshock F b of Kronotskoe earthquake. In the records of stations XAN, COCO and PALK periodic calculations were not revealed. Note that records of stations XAN and COCO were characterized with the noise increased level.

Figure 13
[49]  In the last few days before Hokkaido earthquake of September 2003 with M = 8.3 and the epicenter coordinates [41.81oN-143.91oE] considerable foreshocks ( M>5 ) or remote strong earthquakes were not noted. However the appearance of periodic oscillations was detected. We studied records of stations PET, YSS, OBN, ERM, MAJ, INC, MDJ, and BJT the location of which is shown in Figure 2. From calculations of parameter D it was revealed that at three stations PET, YSS and MDJ oscillations were noted 16 hours before the earthquake, which are presented in spectrum-time diagrams in Figure 13 (interval 4300-5100 minutes). As distinct from Kronotskoe and Sumatra earthquakes, oscillations were revealed in more low-frequency range with periods of 120-160 min. One more group of oscillations was revealed fifty hours before the shock. Similar to Kronotskoe earthquake, periodic oscillations appeared in the records of stations close to the epicenter. Unfortunately ERM station located in the epicentral zone stopped recording 4 day before the earthquake.

Synchronization of Oscillations

[50]  Methods 3 and 4 were used to search for oscillations synchronization effects at stations spaced apart.

Figure 14
[51]  In Figure 14, current values of Hurst generalized index aast are shown, which realizes the maximum of microseismic field singularity spectrum at 5 stations, the location of which is shown in Figure 1. We studied the interval of 30 days before Kronotskoe earthquake (309th-339th days in 1997). We used the window of 12-hour duration with a shift of one hour. With considerable variations of amplitude aast we failed to reliably separate intervals of synchronous oscillations at different stations. But coherence spectral measure calculation l(t, w) as the modulus of canonical
Figure 15
coherences product component by component (formula (9)) allows detecting them (Figure 15). The major burst of coherence is centered in the vicinity of time markers of 40000-42000 minutes (several days before the earthquake). As the moving time window is approaching the earthquake moment, coherence level of aast-variations drops although it remains higher than background statistical fluctuations. Diagrams in Figure 15 testify to the increase of the time duration of low frequency "coherence spot'' as the number of stations increases.

[52]  To compare coherence level with different sets of stations their number should be constant according to formula (9). Sorting all possible combinations of 3 stations lead us to the conclusion that the set of stations PET, MAG, YAK have the highest value of aast = 0.65 and the distance from Kronotskoe earthquake epicenter to those stations is 350, 900 and 2050 km; stations OBN, ARU and YAK have the least value (0.32) and they are the farthest from the earthquake, located at 6800, 5900 and 2050 km. In all variants the coherence measure has a burst in the vicinity of the marker of 40000 minutes, which corresponds to observation time interval of 29.11-03.12.1997 that is 3-7 days before the shock. We note that on 03.12.1997 an intense series of foreshock activation started before Kronotskoe earthquake. Besides in this interval the increase of asymmetric impulses number was observed at PET station, which is the nearest to the earthquake epicenter (Figure 6). Just how random may be these coincidences is difficult to judge.

Figure 16
[53]  In the analysis of the situation before Hokkaido earthquake, the largest number of stations participating in the computation was six: YSS, MDJ, INC, BJT, PET, OBN (Figure 2). Records of stations ERM and MAJ were not used because the former, as it was mentioned above, did not register microseisms in the last four days before the earthquake and the latter was not operated for 7 days two weeks before the earthquake. It was established that synchronization was manifested 2 days before the earthquake (interval 33000-35000 minutes). It encompassed time periods from 3 hours (frequency 0.005 1 min-1 ) and longer ones. As the number of stations decreased, the amplitude l(t, w) increased in accordance with formula (9). It was significant that with complete sorting by 3 stations the most vivid effect was observed for stations nearest to the epicenter of Hokkaido earthquake. Spectrum-time diagram for such stations YSS, MDJ, INC is shown in Figure 16. Three features may be noted: 1) synchronization with period ~3 hours (frequency ~0.005 1 min-1 ) started 9 days before the earthquake (23000 minutes); 2) most vividly and in a wide range of periods it was manifested 2 days before the earthquake (33000-35000 minutes); 3) a break in synchronization in the interval of 29000-31000 minutes was evidently associated with two remote strong earthquakes (shown with arrows) with magnitude 6.6. The first of them with the epicenter coordinates [19.72oN-95.46oE] occurred on 21 September and the second one with coordinates [21.16oN-71.67oW] occurred 10 hours later on 22 September. Of course, the arrival of seismic waves to stations at different times disturbed synchronization.

Figure 17
[54]  In Figure 17, frequency-time diagram l(t, w) is given that was obtained from record processing of stations CHTO, KMI, XAN, COCO before Sumatra earthquake. Beginning with the time marker of 12800 minutes, coherence measure rise occurs with gradual lengthening of prevailing periods of oscillations from several minutes to tens of minutes. In the assumption of intra-terrestrial mechanism of the oscillations, they may be associated with resonance effects in the blocks increasing in scale and/or in the lithosphere layers and deeper layers of the earth. The analysis of microseism amplitude in the second range of periods showed that at all the above-mentioned stations in the interval of 16-26 December their level was practically stationary, which eliminates the atmosphere effects. Such phenomenon was previously noted in both the range of very long periods of the order of 1 year in seismic catalog studies and laboratory experiment of deforming and destructing a sample [Sobolev, 2003]. Apparently it is a fundamental characteristic of non-equilibrium system approaching instability.


Powered by TeXWeb (Win32, v.2.0).