Regional Seismic Phase Arrival Anomalies: GEYOCHA array site.
A. Gashin, L. Haikin, A. Kushnir, M. Rozhkov
SYNAPSE Science Center, 119526Moscow Vernadskogo Ave 101/1 suite 303,
International Institute for Earthquake Prediction Theory and Mathematical Geophysics.
Russian Academy of Sciences.
Warshavskoye sh., 79, korp.2, Moscow556, 113556, Russia.
Tel.: (7 095) 1107795; Fax.: (7 095) 3107032; Email: mitpan@mitp.rssi.ru
Introduction. Caspian basin is a very attractive region for seismic monitoring installations, because it hypothetically covers main areas where violations of the Comprehensive Test Ban Treaty are possible. However, deployment of automatic realtime monitoring system, if ever realized in this region, requires very specific design of the event locator in algorithms. Also, existence of so called oceanic crust in the south Caspian Basin, depressing regional Lg and teleseismic surface waves, creates uncomfortable conditions for seismic event discrimination.
In this paper we present results related to estimation of seismic event parameters: azimuth and apparent velocity using data from Geyocha small aperture array, deployed by IRIS in a vicinity of Ashgabad.
Arrival direction estimation of seismic phases. For estimation of the arrival direction of seismic phases we used different algorithms of spatial spectral analysis: wide band and smoothed low and high resolution algorithms [12] and also generalized 2 and 3 component FK analysis algorithms [3].
For analysis we have chosen two China Lop Nor explosions and regional earthquake. Some characteristics of the events are enclosed in the Table 1.
Table 1. Event characteristics.
Ev. n

Origin time

Source coordinates degr.

Source depth
km.

Magnitude
mb

Epicenter distance degr.

Back azimuth degr.

Pwave appar. veloc.

Source type

1

07.10.94
03:25:58.1

41.66 N
88.75 E

0

6.0

23.74

71.29

11.46

China nuclear
test.

2

10.06.94
06:25:58.0

41.69 N
88.79 E

0

5.7

23.77

71.22

11.46

China nuclear
test.

3

18.10.93
13:57:14.6

22.13 N
62.85 E

10

5.2

16.29

164.13

8.67

Earthquake. Oman Gulf.

Analysis of Geyocha seismograms from the Lop Nor nuclear explosions. The real medium beneath the array has very complex geological characteristics and this seriously hampers the analysis of event phase characteristics. As result, even at the first seconds after Pphase onsets the particle motion of the Pwave has an elliptic feature that is peculiar to heterogeneous and anizotropic media [4]. The results of polarization analysis of the Pwave for different 3C sensors of the array show that deviations of the azimuth and incidence angle estimates relative to their mean value are very high. The mean value of the azimuth estimates is about 22^{0} less then the real azimuth value for first explosion and is 10^{0} less for second one. The mean square errors for the estimates are about 15^{0}. This testifies that serious medium heterogeneity exists even inside the small array aperture (less then 2 Km). The anomaly polarization of the longitudinal waves in the area is previously mentioned in [5].
For the event N1 the FK analysis of the Pphase made using recordings from 12 vertical Geyocha seismometers gave the Pphase arrival azimuth estimate equal 67.6^{0} that is much closer to the theoretical azimuth value. The 4^{0} deviation of the azimuth estimate to the North can be explained by the impact of the great Tibet and TjanShan mountain provinces at the path of the wave propagation. The FK estimate of the Pwave apparent velocity equal to 11.9 km/sec is also well corresponds to the value determined from the JeffreysBullen travel time tables (11.46 km/sec).
For the event N2 the seismograms from only 6 Geyocha sensors were available for the study. The dispositions of these sensors was very unlucky for the FK analysis of signals arriving from the Lop Nor test site. For this reason the FK map of Pwave calculated by conventional algorithm applied to 6 Zcomponent seismograms has a very smoothed maximum. The estimates of wave arrival parameters: azimuth =77.5 and apparent velocity V_{ap}=10.9 km/sec are farther from theoretical values than for the discussed above event N1. The FK estimates made using different time intervals of the Pphase waveform demonstrated rather strong variability. Such instability of the conventional Zcomponent FK analysis stimulated us to employ generalized FK analysis [3] based on 2components (horizontal) and 3component seismograms recorded by the 6sensor subarray. The 2component FK analysis taking into account only transverse phase oscillations provided the azimuth and apparent velocity estimates =69.5, V_{ap}=11.7 km/sec which are significantly closer to the theoretical values than estimates by the conventional Zcomponent FK analysis.
The procedure of generalized 3component FK analysis of Pphase wavefield requires the information about longitudinal wave velocity V_{p} beneath the array. If this information is absent the 3C FK analysis can be made repeatedly with different suspected V_{p} values. A V_{p} value which provides the highest maximum of the FK map can be regarded as the estimate of the real V_{p} value. This procedure applied to the 3C Pwave seismograms of the event N2 gave the V_{p} estimate equal to 4.3 km/sec. The 3C FK map provides the azimuth estimate equal 67.7 which is almost the same as for the FK azimuth estimate got for the event N1. However, the apparent velocity estimate obtained by this method turned to be excessively high: 16.1 km/sec, that is almost in 1.5 times larger then the theoretical value. Since the Pphase polarization even at the first seconds after onset moment has the strongly elliptic feature it is natural to employ the generalized 3C FK analysis for the elliptic particle motion polarization model. The Pphase azimuth and apparent velocity estimates got with the help of this algorithm are equal to =71.7^{0}, V_{ap}=12.1 km/sec and are the most close to the theoretical values =71.2^{0}, V_{ap}=11.5 km/sec, than all previously mentioned estimates.
The comparison of the results of polarization and FK analysis allows estimate the Pwave velocity in the medium beneath the array. Employment of the simplest relation V_{p}=V_{ap}sin (where is the incidence angle) gives the value V_{p}=3.6 km/sec for the first event and V_{p}=3.7 km/sec for the second one, that is significantly less than value V_{p}=5.5 km/sec assumed in the JeffreysBullen Earth model but corresponds to results of the medium structure assessment by the receiver function analysis [5].
After filtering made in the frequency band 0.010.1 Hz intensive Rayleigh and Love phase oscillations are explicitly seen at the event seismograms. This gave the chance to estimate the arrival parameters of surface waves with the help of generalized 3component FK analysis. This analysis provided the following estimates for the event N2  for the Rayleigh wave: =66.1 and V_{ap}=2.2 km/sec; for the Love wave: =64.1 and V_{ap}=3.3 km/sec. The azimuth estimates for the both waves differ from theoretical value equal to 71.2. This undoubtedly connected with the impact of Pamir and TjanShan mountains which lie at the theoretical path of wave propagation. The same reason probably explains the low apparent velocities of the surface waves. The azimuth and apparent velocity estimates for different phases of N1, N2 events are enclosed in Tables 2a, 2b.
Analysis of Geyocha 3C seismograms from the Oman gulf earthquake. In seismogram of this event we can see three phases of body waves (Pn, Pg and Sn). In accordance with theoretical prediction the Pn phase has the most high frequency content: its spectrum maximum is at 0.7 Hz (but there exists the second powerful spectral peak with frequency 0.4); the Pg phase has the single spectral peak at 0.55 Hz, and the Sn phase  at 0.33 Hz. The FK analysis maps for these phases are presented in Fig.1a1c. The estimates of Pn and Sn waves azimuths are equal 184.7 and 183.8, i.e. coincides with the accuracy less than 1 but differ from theoretical value (equal to =164.1) of about 20. The apparent velocities of this phases are V_{ap}(Pn)=8.7 km/sec and V_{ap}(Sn)=8.0 km/sec that are close to the theoretical values. The azimuth divergence of the Pn and Sn phases can not be explained by the estimation errors and obviously connected with the peculiarities of this wave propagation in the region. Both phases propagate along the boundary between the Earth crust and upper mantle and probably this boundary has a laterally heterogeneous structure in the region.
At the same time the Pgphase FK azimuth estimate is equal to 169.1 and differ from the theoretical azimuth only in 5. Because the Pg phase propagates within the Earth crust the mentioned lateral heterogeneity of the crustmantle boundary apparently does not affect strongly on this wave propagation. This matter is already discussed in the papers [5, 6] where was stated that various wave phases demonstrate the different arrival azimuths in Central Asia region. However, the authors of these papers observed this effect only for the surface waves.
Fig.2a shows the Love wave power spectral density after filtering made for frequency band 0.012 Hz. The spectrum peak is focused in frequency band 0.0250.09 Hz with maximum at 0.05 Hz. The power spectral density of the Rayleigh wave longitudinal oscillations is shown in Fig.2b. We see that the Rayleigh wave spectrum has the rather narrow peak concentrated in frequency band 0.050.1 Hz. with maximum at 0.08 Hz
The wideband FK analysis of the Love wave seismograms accomplished for frequency band 0.030.08 Hz. revealed that the arrival azimuth and apparent velocity of the Love wave are equal to =157.6^{0} and V_{ap}=3 km/sec. The azimuth of the Love wave arrival is less than theoretical one for 6,5^{0}, and demonstrates the very high consistency in different frequency bands. This follows from the results of the multiple narrow band FK analysis presented in Fig.3a and Fig.3b The FK analysis was performed for 6 equidistant frequency bands with width equal 0.1 Hz in the range 0.030.08 Hz. From Fig.3a one can see that the Love wave arrival azimuth does not practically change depending on the wave period. This contradicts with the results of paper [6] where such dependence were mentioned (however for the other arrival direction of surface waves). Fig.3b presents the dispersion curve of Love main mode group velocity. This dispersion curve good corresponds to the one presented in [5].
The wideband FK analysis of Rayleigh wave was made in frequency band 0.050.1 Hz. The arrival azimuth estimate is equal 168.8^{o}, that exceeds the theoretical value at 4.7^{o} and the arrival azimuth estimate for the Love wave at 11.2^{o}. Note that the theoretical azimuth is almost in the middle between the above surface waves azimuths. Such great difference between the Love and Rayleigh azimuth estimates can not be explained by the errors of the FK analysis if to take into account the very high signalto noise ratio in the both phase seismograms.
Conclusions. This study has shown that the south Caspian has an anomalous crustal structure which has a pronounced effect on not only higher frequency regional seismic waveforms but also on lower frequency surface waves. Azimuth estimations based on data from a 3component station or small aperture array provides the support for automated near real time event location based on observations from a single site. The results of azimuth estimation study made for recordings of Geyocha 3component very broad band array show that in the regions with complex medium structure as the south Caspian Basin azimuth measurements can be misleading and potentially provide the gross errors of epicenter location.
The azimuth estimations by polarization analysis of 3component seismograms exhibit great variability even for station distanced at 12 km. The FK analysis of array data guarantees more precise azimuth estimates, especially while using generalized 2 and 3component FK algorithms. Nevertheless even FK azimuth estimates reveals significant difference in azimuths of body and surface event wave phases arriving to Geyocha site from Eastern and Southern directions. The azimuth and apparent velocity estimates for different phases of N3 event are enclosed in Table 2c.
Acknowledgments: The authors thank Dr. Anton Dainty from Phillips Lab. (USA) provided data sets from the Geyocha array.
Study was funded by USAF EOARD and by ISTC (project N451).
References.

Capon J., 1969, Highresolution frequencywavenumber spectrum analysis, Proc. IEEE, v.57, pp.14081418.

Kushnir A.F., 1995, Algorithms for adaptive statistical processing of seismic array data, Husebye E. and A. Dainty (eds.) Monitoring of Comprehensive test Ban Treaty, Kluwer Academic Publishers, Dortrecht, Boston, London, pp.565586.

Kushnir A.F. , 1997, Estimation of apparent slowness vector for a plane wave using data from a threecomponent seismic array: a statistical problem involving nuisance parameters, Computational Seismology, v.29 (in press).

Der, Z.A., Baumgardt, D.R., Shumway, R.H., 1993, The nature of particle motion in regional seismograms and its utilization for phase identification, Geophys.J.Int., 115, pp.10121024.

Mangino S., K. Priestley, 1995, Velocity Structure in the Region of the South Caspian Basin from Teleseismic Receiver Function Modeling. Scientific Report, Phillips Laboratory, Hanscom AFB, MA 017313010.

Pavlis, G., Mahdi, H., Vernon, F.L., 1994, Surface wave propagation in Central Asia: Observations of scattering and multipathing with the Kyrgyz broadband array, Proc. of the 16th Annual Seismic Research Symposium, pp.291297.
