|Observatories and Research Facilities for EUropean Seismology|
|Volume 3, no 2||December 2001||Orfeus Newsletter|
Focal properties of the MW=6.5 Skyros, Aegean Sea, earthquakeN.S. Melis1, G.N. Stavrakakis1, J. Zahradnik2
The earthquake and its aftershock sequence have been recorded by the Greek National Network and were located by NOA (National Observatory of Athens). NOA also determined the earthquake magnitudes; the main event had a surface wave magnitude MS=5.8 and a local magnitude ML=5.3. After the main shock, NOA quickly assembled information on the NOA webpage for the interested public.
The purpose of this study is to understand the basic focal properties of the Skyros main shock using newly available broadband data from the Greek National Network. Another goal of our study is to demonstrate rapid investigation and result dissemination capabilities of an ongoing seismic sequence, utilizing readily available regional network data; these capabilities are useful for the PRESAP project (EU project - EVG1-CT-1999-00001).
Figure 1. Distribution of epicentres of the Skyros aftershock sequence. Shown are events with ML=3.5 that occurred less than two months after the main event. Focal mechanisms from first motion data are shown for the main event, two foreshocks and the strongest aftershock.
Lennartz sensors. One additional station SER (Sergoula), is a stand-alone station equipped with a 100s CMG3-T Guralp sensor. Specific webpages of NOA and Charles University include the exact station co-ordinates and allow a down-load of the most significant SER records for the present sequence.
Figure 2. Broadband stations recorded the Skyros event in Greece. With red triangle NOA stations, green triangle SER (Sergoula) station. Star denotes the main shock epicentre.
All records are instrument corrected, re-sampled to time increment of 0.02 sec, high-pass filtered (frequencies f > 0.05 Hz), and rotated into R (radial), T (transversal), Z (vertical), and integrated to displacement. The first-motion polarities, carefully read from three component seismograms at all stations provide additional constraints on the focal mechanism. Projecting polarities on the focal sphere is a delicate problem (Zahradnik et al., 2001). To avoid unrealistic take-off angles of the first arrivals, formally interpreted as head waves from inter-crustal discontinuities, we use a gradient model GMF. GMF is an approximation of the layered model MF, recently obtained for travel paths from northwestern Turkey to Greece by inversion of Love wave dispersion (Novotny et al., 2001), see Figure 3. The take-off angles in the gradient model were calculated with the ray-method code ANGGRA (Jansky, 2001).
Figure 3. Crustal models used in this study. The gradient model GMF is an approximation of the homogeneous layer model MF. Both models have the same Moho discontinuity at 33km depth.
With the ASPO method, we analyze the displacement amplitude spectra of complete 3-component waveforms (duration of 160 seconds), in the frequency range 0.05-0.08 Hz (below the corner frequency). The observed spectra are compared to synthetic spectra calculated with the discrete-wavenumber method (Bouchon, 1981; Coutant, 1989) using model MF. For a set of trial source depths, we perform a systematic 10 degrees grid search for the strike (0o-360o), dip (0o -90o), and rake (0o-180o), that best fit the synthetic spectra. A grid search 0o-180o degrees for the rake is sufficient, since solutions with rake R and R-180o have the same amplitude. Scalar moment affecting the spectra linearly is not searched, but estimated from the ratio between the observed spectra and the unit-moment synthetic spectra. For details, see Zahradnik (submitted). The best fitted focal mechanism is then used to calculate the synthetic seismograms.
We started with all nine stations, but the synthetic seismograms could not fit the nearest station with a very good signal-to-noise ratio, ATH (133 km). Then we tested several station sub-sets and the best results (unique misfit minimum and good fit to ATH station) were obtained for the amplitude-spectra inversion from three stations: ATH (133 km, azimuth 205o), PRK (167 km, 83o) and APE (244 km, 155o).
The misfit between the observed and synthetic amplitude spectra (sum from all stations, components, and frequencies) for the three NOA stations is plotted against the sequential number of the strike-dip-rake trial (Figure 4). The least misfit values that range from the minimum to 1.05 times the minimum misfit - the range is used to measure solution uncertainty - are marked by blue crosses. Only two minima fall in the error range, one with strike = 150o, dip = 70o, rake = 10o, and its conjugated solution 57o, 81o, 160o. Such a unique solution is quite an exception compared to ASPO applications for other earthquakes; the present solution is very well constrained by the available data. The ASPO method also compares the observed and calculated first-motion polarities. The red diamonds in Figure 4 mark the fault-plane solutions that are consistent with all nine P-polarities and, at the same time, have a spectral amplitude misfit between min and 1.05*min. The amplitude-preferred solution and the first motion polarities are entirely consistent (coincidence of the blue crosses and red diamonds), which is again a very rare case.
Figure 4. Misfit function of the grid-search ASPO modeling for the mainshock. The trial number on the horizontal axis refers to the sequential number of the systematic search (triple loop) over the strike, dip and rake.
The result in Figure 4 is for a focal depth of 8 km, which (together with a depth of 9 km) has the lowest amplitude-spectra misfit of the focal depths tested, see Figure 5.
Figure 5. Variation of the amplitude misfit with the focal depth for the mainshock.
For a focal depth of 8 km, and the above focal mechanism, the ratio between the observed amplitude spectra and the unit-moment synthetic spectra (averaged over the frequency range 0.05-0.08 Hz) yields the scalar seismic moment of M0 = 4.1·1018 Nm (corresponding to moment magnitude MW = 6.5). Our final result, strike = 150o, dip = 70o, rake = 10o, M0 = 4.1·1018 Nm, obtained from regional data, is in good agreement with moment tensor solutions determined with teleseismic (USGS strike = 145o, dip = 85o, rake = 4o, M0 = 5.4·1018 Nm and Harvard strike = 148o, dip = 71o, rake = -1o, M0 = 5.7·1018 Nm) and regional data of relatively distant (mostly > 1000 km) stations (Swiss Seismological Service strike = 148o, dip = 73o, rake = 0o, M0 = 8.7·1018 Nm). Our final result is also close to our fast preliminary determination from eight NOA broad-band stations and 17 NOA polarities, which provided strike = 170o, dip = 70o, rake = 20o, M0 = 4·1018 Nm (NOA webpage). Solving the forward problem and comparing the synthetic and observed displacement waveforms, we find that the final solution provides a better fit than the preliminary one (Figure 6). The main problem of the preliminary solution was its failure to explain the high-quality ATH record. The preliminary solution reverted the 'sign' of ATH's prominent wave group. As seen in Figure 6 the final solution is not ideal, in particular for stations VLS, KZN, RDO, which implies the need for further refinement of the crustal model.
Figure 6. Band-pass filtered displacement transverse components (blue), compared to synthetics of the preliminary (black) and final (red) ASPO fault-plane solution. For some stations the black and red curves coincide with each other. Numbers to the right of the traces indicate their peak values (in m).
http://seis30.karlov.mff.cuni.cz). GMT (Wessel and Smith, 1995) was used to produce some of the diagrams.