IC/2001/72 United Nations Educational Scientific and Cultural
Transcription
IC/2001/72 United Nations Educational Scientific and Cultural
IC/2001/72 United Nations Educational Scientific and Cultural Organization and International Atomic Energy Agency THE ABDUS SALAM INTERNATIONAL CENTRE FOR THEORETICAL PHYSICS SOURCE SCALING FOR THE INTERMEDIATE-DEPTH VRANCEA EARTHQUAKES A. Gusev1 Institute of Volcanic Geology and Geochemistry, 9 Piip Blvd, 683006 Petro-Kamchatskii, Russian Federation, M. Radulian2, M. Rizescu2 National Institute for Earth Physics, P.O. Box MG-2, 76900 Bucharest, Romania and G. F. Panza3 University of Trieste, Department of Earth Sciences, Trieste, Italy and The Abdus Salam International Centre for Theoretical Physics Trieste, Italy. MIRAMARE - TRIESTE July 2001 1 gusev@emsd.iks.ru mircea@infp.ro; Mihaela.Rizescu@infp.ro 3 panza@dst.univ.trieste.it Abstract Earthquake source scaling properties are studied for the intermediate-depth seismicity nest in Vrancea region, Romania, known as a source of many destructive earthquakes. Some earlier work indicated high stress drop for Vrancea earthquakes, and our purpose is to check these findings using wide-band digital records from events of an adequate magnitude range. We could find body-wave records with acceptable signal/noise ratio for sixteen Vrancea earthquakes (3.7 < Mw^ 7.4) which occurred since 1976. Recording stations used are local, regional and teleseismic. To analyze the spectral scaling properties, the common spectral features of the corner frequency, fc, and the spectral slope, y, have been determined for P- and S-waves. This was done, both for entire P or S groups and for separate phases like P or pP or sP, etc. The time-domain durations, T, have been determined from the deconvolved displacement pulses, and the time-domain fc, equal to T 1 , has been estimated. In all the variants of the processing - entire group or single pulses, P- or S-waves, spectral domain or time domain - quite consistent and comparable results have been obtained. The only systematic difference is between P-wave based and S-wave based estimates, represented by the usual fcp/fcs ratio of about 1.4. To determine the scaling of fc vs seismic moment, Mo, we used the Mo values given in the ROMPLUS catalogue, which are consistent with our estimated spectral levels. When viewed over the entire magnitude range (M\y = 3.7 -s- 7.4), the fc vs Mw trend generally agrees with the model of constant stress drop (independent of magnitude). However, some individual earthquakes of about magnitude 5 show very long durations on GRF; this may be a peculiar propagation effect for the path to this station. The value of the constant stress drop, based on the log-average fcs vs Mw trend, is near to 3 MPa. This means that the scaling of intermediate Vrancea earthquakes over the wide magnitude range is quite similar to that normally observed for shallow events, with the typical stress drops range of 110 MPa. However, at the upper-magnitude end, this normal trend is violated. The largest intermediate-depth events (Mw > 6.5) show a clear tendency for higher static stress drops than the shallow events, and for magnitudes M w > 7, stress drops in excess of 10 MPa can be expected. Unfortunately, due to the very limited data volume, this conclusion is not definitive. However, if true, our result means that when one wishes to estimate the strongmotion amplitudes of the largest intermediate Vrancea earthquakes, one cannot merely use the extrapolation of the strong-motion parameters observed for moderate-magnitude events to the largest ones on the basis of the standard constant-stress-drop scaling assumption. This simple approach would lead to significantly underestimated amplitudes. Introduction The Vrancea zone, located in Romania, at the sharp bend of the South-Eastern Carpathians (Figure 1) is a well-defined seismic region in Europe, with unique properties. The seismicity is concentrated in an extremely narrow, high-velocity, focal volume in the depth range from 60 to 200 km. A relatively high seismic energy is persistently released (four shocks with magnitude greater than 7 occurred during the past century) by a seismogenic process still far from being fully understood. At more shallow levels (0-60 km) the seismicity here is sporadic and weak (magnitude below 5.5), and seems to be decoupled from the seismic activity in the subcrustal lithospheric slab. All the major shocks are characterized by a quite stable reverse faulting focal mechanism with the rupture plane oriented in a NE-SW direction, parallel to the Carpathians arc, and following the orientation of the epicenters distribution. The analysis of local short-period data indicates that, for a given seismic moment, the source area of the intermediate-depth Vrancea events is significantly smaller than the size usually observed for shallow events, and implicitly the stress drop is higher (Oncescu, 1989; Radulian and Popa, 1996; Popa and Radulian, 2000). The scaling properties of the source spectrum are extremely important in seismic hazard studies, because they may significantly modify the strong motion parameters as compared to the case of "average" stress drop. For a given magnitude and distance, unusually high stress drop increases acceleration and velocity amplitudes and decreases the duration of shaking; the effect of low stress drop is the opposite. As a result, the assumed values of stress drop may have a drastic effect on the seismic hazard values as computed by the deterministic approach (e.g. Radulian et al., 2000). Thus, it is of paramount importance to determine reliable source spectral scaling. The main goal of this paper is to analyze the scaling properties of the Vrancea intermediate-depth earthquakes and to check if the anomalous scaling suggested by the local short-period data is confirmed by the broadband data recorded at local and global scale. To this aim, we analyze the set of ninety-three broadband records for sixteen moderate and large Vrancea intermediate-depth earthquakes. Observational data We consider the set of broadband waveforms recorded at teleseismic, regional and local distances, for sixteen intermediate-depth Vrancea earthquakes which occurred between 1976 and 2000 (3.7 < M w < 7.4) (Table 1). The use of the data from the station GRF (Grafenberg, Germany), affiliated to the German Regional Seismic Network, with the longest recording history and located at a relatively small epicentral distance (about 11°), has been pivotal for the success of this study. Also highly valuable were the data from the GEOFON station MLR (Muntele Rosu, Romania), operated jointly by the GeoForschungsZentrum Potsdam and the National Institute for Earth Physics, Bucharest, located practically in the epicentral zone of Vrancea earthquakes. Additional data are from representative regional and teleseismic broadband stations with a reasonable azimuthal coverage for Vrancea earthquakes. We selected stations situated inside continents to minimize storm/surf microseismic noise. The following broadband stations (situated at epicentral distances as given in parenthesis) have been used as well: TRI (9°), AQU (10°), OBN (11°), ARU(22°), BGCA (41°), HIA (59°), CCM (80°), ANMO (89°). The set of waveforms has been retrieved from the World Wide Web pages of the IRIS Data Management Center in Seattle, USA and GEOFON Data Center in Potsdam, Germany. Estimation of source parameters For the events listed in Table 1 all usable P- and S-wave records have been processed in order to estimate the source parameters, both in frequency and time domain. A computation algorithm, programmed using the MATLAB software package of Math Works, has been set up to determine interactively the source parameters. The observed velocity data, corrected for instrument response, are integrated and differentiated to get the displacement and acceleration time series. The waveforms are plotted together with the theoretical phases and arrival times, based on the travel times calculated for the IASP91 Earth model (Buland and Chapman, 1983). The records of body waves are corrected for attenuation and geometrical spreading. Attenuation affects both the shape and the spectral properties of the P- and S-wave pulses, -nft* therefore to correct for attenuation we consider the attenuation operator € , where/is the frequency and t* is the travel time, divided by the quality factor. For the P-wave we assume that t* varies in the range 0.6 -1.0 s for teleseismic distances (A>20°), and that it is equal to 0.3 s and 0.05 s for regional (A around 10°) and local distances (A around 0.5°), respectively. For the S-wave the assumed t* value varies from 2.4 s to 4.0 s for teleseismic distances, and it is equal to 0.6 s and 1.0 s for regional and local distances, respectively. The geometrical spreading coefficients, as given by relation (Aki and Richards, 1980): cos iv cos L sin A A 5 P (1) dp (where x and ^ are the position vectors of the station and focus, respectively, v(£) is the seismic wave velocity at the focus, ix and i^ are the incidence angle and take-off angle, respectively, A is the epicentral distance and p is the ray parameter) have been calculated in an isotropic spherical earth model. The common "ocfY" spectral model (Brune, 1970, Hanks and Wyss, 1972), that yields the low-frequency spectral level, the corner frequency fc and the high-frequency spectral decay y have been used to characterize the spectral shape. For each phase, the signal and noise time windows are interactively selected, then a 5% cosine tapering is applied and the displacement spectrum is computed using a FFT algorithm. The signal is noise-corrected by subtracting the estimated noise contribution from the squared spectrum. The spectrum of the signal is inverse-filtered with the attenuation operator, corrected for the geometrical spreading factor and scaled to the seismic moment units. The zero-frequency level, the corner frequency and the slope of the high-frequency decay are interactively determined after visual inspection of the corrected spectrum. The first spectral parameter is an estimate of the seismic moment Mo and provides the moment magnitude by the formula (Kanamori, 1977): Mw=|aogM0[N.m]-9.1) (2) Only the interactive mode allowed us to process records with low signal-to-noise ratio, with 0.2-0.3 Hz microseismic spectral peak just in the middle of the analyzed frequency band. A few examples of the analyzed waveforms with high and low signal-to-noise ratios are given in Figure 2. Signal and noise spectra are also plotted. The spectral shape has been estimated on all usable components and separately for P, pP, sP, S and sS phases, whenever they could be unequivocally identified. Our tests show that the spectra obtained when considering isolated single wave pulses or wave groups (containing P, pP and sP, for example) are consistent with each other, both as corner frequency and spectral slope. The values obtained from the different components of motion also show a good consistency. Therefore, the seismic moment, corner frequency and spectral slope for P- and S-waves have been estimated as averages of all the reliable values determined from the different phases and components of motion (Table 2). The correction for attenuation is performed, in the frequency domain, on the displacement spectrum. To calculate the corresponding phase correction, that follows from the requirement of causality, a program module from Claerbout (1976) was used. The pulse duration is interactively estimated in the time domain, where the pulse is low-pass filtered, with a cutoff frequency of 2-4 Hz (typical for the teleseismic case), using a 4 th order Butterworth filter, to remove the high-frequency noise produced by the normalization with the attenuation operator. For the time-domain analysis, only isolated P- and S-phases have been considered, and the corresponding results are given in Table 2. Only a few estimates have been published for the corner frequencies and source durations of Vrancea intermediate depth earthquakes. From the broadband recordings of the 1977 event, Fuchs et al. (1979) give, for S-waves, fc = 0.07 Hz, in agreement with our results. The source durations for the strong events of 1977, 1986 and 1990 (first and second shocks) determined from teleseismic waveforms inversion (see Tavera (1991) for references) are also compatible with our determinations. From the strong motion data of the 1986 event Oncescu (1989) gives, for S-waves,/ c = 0.23 Hz. This value, relatively high if compared with ours, can be explained by the fact that only the asperity, and not the whole source area, is seen by the accelerogram record. The difference is natural because analogue strong motion data cannot provide a good coverage of the low-frequency part of the spectrum. The computation of the source parameters for small to moderate Vrancea events, using local data (Oncescu, 1986; Rizescu, 1999; Radulian and Popa, 1996), gives values of/c similar to the ones we obtain for the MLR station. The comparison of the Mo and M w values that we have determined, from the zerofrequency spectral levels, with the corresponding values given by the Romanian earthquake catalogue (Figure 3) shows, in general, an acceptable agreement, with no systematic bias. We prefer to use moment magnitudes (Mw) from the regional catalogue, since we consider these estimates, based on all available data for each event, more reliable than our spectral estimates, based, in many cases, on the data of a single station. Scaling relationships and their discussion We now try to establish the relationship fc vs M w that defines the source scaling. The observed Mw -fc relation is represented in Figure 4 (for P-waves) and Figure 5 (for S-waves). Linear trends that represent the characteristic constant-stress-drop average behavior, following Brune (1970), are also plotted. For S-waves: log/ cS = -0.5M w +C s (3) where Cs=2.25 or 2.6 for 1 or 10 MPa stress drop, respectively. The relationship of this kind was first established by Thatcher and Hanks (1973) as the average empirical trend for moderate-to-large shallow California earthquakes. Since that paper, a lot of work has been done following Brune's spectral approach, and for practically all studied sets of moderate or moderate-to-large shallow earthquakes, (3) holds at least approximately. However, the values of the average stress drop vary significantly, covering the range 0.5 - 10 MPa. A certain part of this scatter probably reflects some differences in processing and analysis procedures; another contribution is the natural dispersion between various data sets. To derive the analogue of (3) for P-waves, we assume the typical empirical value of the corner frequency ratio: f(P/fcs=l-4 (4) and obtain: (5) where Cp=2.4 or 2.75 for 1 or 10 MPA stress drop, respectively. The coefficient 0.5 in relations (3) and (5) expresses the standard, constant stress drop source scaling. Assuming this kind of scaling, we approximate our data points with a linear relation with the fixed slope value equal to -0.5, and with the intercept value obtained from the bestfit (dashed lines in Figures 4 and 5). The corresponding values of the stress drops for P- and S-waves are 3.1 MPa and 3.2 MPa, respectively. These values of stress drop are within the interval usually observed for shallow earthquakes when Brune's approach is used. They are also close to the characteristic values for the inter-plate earthquakes after Kanamori and Anderson (1975). Though this result is evidently a good first approximation, some deviations are seen at the upper-magnitude end, and will be discussed later. Another interesting feature is seen in the magnitude range from 5 to 5.5, where the dispersion of our estimates seems to be unusually high. The time domain estimates behave in a quite similar way. In fact the (1/T) vs Mo relationships, plotted in Figures 6 (P-wave) and 7 (S-wave), are very close to the fc vs M w relationships of Figures 4 and 5. This consistency is also visible when looking at the correlation between fc and the source duration (Figures 8 and 9). The/ C and the inverse of the source duration closely follow the line with the slope 1, only slightly shifted towards smaller 1/T values. On the average, fcP/(l/TP)=1.0 and f cS /(l/T s )=l.l. Thus, the scaling revealed in the first approximation for Vrancea intermediate-depth earthquakes follows the general trend of shallow earthquakes. The ratio of the corner frequency values, derived from P- and S-waves is close, on the average, to the value of 1.4, quite typical for shallow earthquakes. The relationship between the inverse pulse durations for P and S-waves is similar. The dispersion of the values obtained with the analysis in the frequency domain and in the time domain, is comparable both for P- and S-waves (Figures 4 and 6, 5 and 7, respectively). The good match between the results of frequency-domain and time-domain analyses adds an additional weight to our general conclusions. When viewed over the entire earthquake size range (Mw = 3.7 -s- 7.4), the observed fc vs Mw trend generally agrees with the model of constant stress drop (independent of magnitude). However, at the larger-magnitude end, this model becomes invalid. The largest events (Mw > 6.5) show a clear tendency for stress drops that are larger than those for moderate-magnitude ones (or, for typical shallow events). From the visual inspection of. Figures 4 and 5, one can expect, for Mw > 7, stress drop values as high as 10-20 MPa. This possibility is very important for prediction of future strong motion, because at a given Mw, increased stress drop produces larger peak velocity and acceleration values and response spectra, and smaller durations. The problem of the existence of a simple and theoretically plausible scaling law is a general one, and has very important consequences on seismic hazard estimation issues. Theoretical source models usually assume constant stress drop and constant rupture velocity, resulting in a simple fc vs Mw scaling, like the one described by equations (3) and (5). However, natural earthquakes need not follow these or other similar assumptions. In the case of Mexican earthquakes (Singh et al 1989), it was found that when strong motion amplitudes are extrapolated from moderate to large earthquakes, according to the constant stress drop model, the resulting estimates are in excess with respect to the values actually observed. The situation in the case of the Romanian earthquakes seems to be the opposite: the extrapolation from moderate earthquakes would rather underestimate the strong-motion amplitudes of the largest events. In such a situation, using empirical and not "theoretically-based" scaling may lead to more reliable results. One such case is the study of Moldoveanu and Panza (1999) who, using a kind of empirical source scaling modelled successfully the main characteristics of the accelerogram recorded in Bucharest for the strong 1990 earthquake (Mw=6.9). In some cases, the S-waveform at GRF station is unipolar, whereas the P-waveform of the same event is of comparable duration but bipolar (approximately doubling the value of/ c ), as seen in the example in the center of Figure 2. Quite hypothetically, this fact may be explained by a difference between the propagation mode of P- and S-waves, around 11° along this particular path, with a single branch for S-wave and multiple branches for P-wave. The distributions of the estimated P-wave and S-wave spectral slopes as functions of magnitude are plotted in Figures 10 and 11, respectively. The /values for S-wave are in the range 1.8 - 2.0, while the / values for P-wave are generally larger, with a tendency to increase with increasing earthquake size. This tendency, not clearly defined, is however hardly an artifact due to the approximations in our procedure, and deserves some discussion. Our /estimates are in fact somewhat questionable because (1) true t values for the actual paths are unknown and (2) we assume that t is frequency dependent and this model is too simple. The first factor can hardly cause the dependence of / on magnitude, while the second factor may provide an explanation. Indeed, our analysis suggests that the true teleseismic t may significantly decrease between 1 and 6 Hz, reaching values as low as 0.1 0.2 s at 5 - 6 Hz. In such a case, our correction for attenuation is poor and our /estimates are based on somewhat biased, too gradually sloping spectra. The magnitude of this bias depends on the actual frequency band in which the data are analyzed that, in turn, depends on the signal-to-noise ratio. At the largest magnitudes, the utilizable frequency band is the widest, and the mentioned bias is the largest as well, thus leading to an underestimate of /. Therefore, when we are ignoring the frequency dependence of t* we must introduce an artificial decrease, with increasing magnitude, of the /estimates. The actual tendency is the opposite one. This may indicate that the increase, with increasing magnitude, of the value of /for the P-wave spectra is a genuine phenomenon. The lack of a similar tendency in S-wave is not too relevant because the S-wave spectra of the larger events are teleseismic and are thus of relatively lower reliability. Conclusions The scaling relations for the Vrancea subcrustal earthquakes have been analysed using both frequency- and time-domain measurements from broadband recordings. The data set is formed by sixteen events which occurred between 1976 and 2000 (3.7 < Mw^ 7.4), recorded by ten stations spanning the epicentral distance range from 0.5° to 90°. The two kinds of measurements (frequency- and time-domain) are well correlated over the whole magnitude interval considered, and the inverse of the pulse duration is a good estimate of the corner frequency. The ratio of the corner frequency values, as well as the ratio of the inverse pulse duration values, deduced from P-and S-waves, are close to the value 1.4, a typical one for crustal events. The earthquake size scaling with corner frequency or source duration follows, in the first approximation, the general trend observed for shallow earthquakes. For the whole magnitude range (Mw between 3.7 and 7.4) our data distribution is well approximated by a linear dependence with slope 0.5, typical for crustal earthquakes. The large dispersion in the 5-5.5 magnitude interval is partly explained by the anomalously low fc and 1/T values obtained with Grafenberg station (GRF). This station is the only one that provides observations for a broad magnitude range, including the largest magnitudes, so it is crucial for our study, but unfortunately strong path effects might contaminate its recordings of Vrancea earthquakes. For the largest magnitudes range (Mw > 6.5) our study evidences a deviation of the source scaling towards larger stress drops. This result has significant consequences on seismic hazard estimation, due to Vrancea intermediate-depth earthquakes. It may mean that the empirical scaling of strong motion parameters may in certain cases be closer to reality than the "theoretically-based" ones that use physically sound but in fact too simplistic assumptions. The applicability of a simple scaling is further questioned by some indication that the spectral slope increases with increasing magnitude. Unfortunately, these very important results should be considered as preliminary ones because of the limited data set, which is available. Acknowledgements This research is a contribution to the NATO SfP project 972266 "Impact of Vrancea Earthquakes on the Security of Bucharest and other Adjacent Urban Areas (Ground Motion Modelling and Intermediate-Term Prediction)" and to the UNESCO-IUGS-IGCP project 414 "Realistic Modelling of Seismic Input for Megacities and Large Urban Areas". We used the Tau software to calculate travel times and SPHERAY software to compute geometrical spreading coefficients, available on Internet at www.orfeus.knmi.nl and www.geoscope.ipgp.jussieu.fr, respectively. References Aki, K., Richards, P.G., 1980. Quantitative seismology. 2 volumes, W.H. Freeman, San Francisco, 932p. Brune, J.N., 1970. Tectonic stress and the spectra of seismic shear waves from earthquakes. J. Geophys. Res., 75, 4997-5009. Buland, R., Chapman, C.H., 1983. The Computation of Seismic Travel Times, Bull. Seism. Soc. Am., v. 73, pp. 1271-1302. Claerbout, J. F., 1976, Fundamentals of Geophysical Data Processing, McGraw-Hill, N.Y Fuchs, K., Bonjer, K.-P., Bock, G., Cornea, I., Radu, C , Enescu, D., Jianu, D., Nourescu, A., Merkler, G., Moldoveanu, T., Tudorache, G., 1979. The Romanian earthquake of March 4, 1977 II. Aftershocks and migration of seismic activity, Tectonophysics, 53, 225-247. Hanks, T.C., Wyss, M., 1972. The use of body-wave spectra in the determination of seismicsource parameters. Bull. Seism. Soc. Am., 62, 561-589. Kanamori, H., 1977. The energy release in great earthquakes. J. Geophys. Res., 82, 29812987. Kanamori, H., Anderson, D.L., 1975. Theoretical basis of some empirical relations in seismology, Bull. Seism. Soc. Am., 65, 1073-1095. Moldoveanu, C.L., Panza, G.F., 1999. Modelling, for microzonation purposes, of the seismic ground motion in Bucharest due to the Vrancea earthquake of May 30. In: Vrancea Earthquakes: Tectonics, Hazard and Risk Mitigation, F. Wenzel, D. Lungu (eds) & O. Novak (co-ed.), 85-98, Kluwer Academic Publishers, Dordrecht, Netherlands. Oncescu, M.C., Marza, V.I., Rizescu, M., Popa, M., 1999. The Romanian earthquake catalogue between 984-1997, in: Vrancea Earthquakes: Tectonics, Hazard and Risk Mitigation, F. Wenzel, D. Lungu (eds) & O. Novak (co-ed.), 43-47, Kluwer Academic Publishers, Dordrecht, Netherlands. Oncescu, M.C., 1986. Some source and medium properties of the Vrancea seismic region, Romania. Tectonophysics, 126, 245-258. Oncescu, M.C., 1989. Investigation of a high stress drop earthquake on August 30, 1986 in the Vrancea region. Tectonophysics 163, 35-43. Popa, M., Radulian, M., 2000. Test of the empirical Green's function deconvolution on Vrancea (Romania) subcrustal earthquakes. Studia Geoph. et Geod. 44, 403-429. Radulian, M., Popa, M., 1996. Scaling of the source parameters for the Vrancea intermediate depth earthquakes. Tectonophysics, 261, 67-81. Radulian, M., Vaccari, F., Mandrescu, N, Panza, G.F., Moldoveanu, C.L., 2000. Seismic hazard of Romania: deterministic approach. Pure appl. geophys. 157, 221-247. Rizescu, M., 1999. A completely automated system for seismological data acquisition, processing and exchange, PhD Thesis, Institute for Atomic Physics, Bucuresti, 219 p. Singh, S.K., Ordaz, M., Anderson, J.G., Rodriguez, M., Quaas, R., Mena, E., Ottaviani, M. & Almora, D., 1989. Analysis of near source strong-motion recordings along the Mexican subduction zone, Bull, seism. Soc. Am., 79, 1697-1717. Tavera, J., 1991. Etude des mecanismes focaux de gros seismes et seismicite dans la region de Vrancea, Roumanie, Raport de stage de recherche. Thatcher, W., Hanks, T.C., 1973. Source parameters of Southern California earthquakes. J. Geophys. Res., Vol 78, 8547-8576. Table 1. Selected intermediate-depth Vrancea earthquakes. Locations and magnitudes are taken from the Romanian Earthquake Catalog ROMPLUS (Oncescu et al., 1999), where Mw values are moment magnitudes from teleseismic observations for Mw > 6.0 and moment magnitudes or duration magnitudes, from short-period recordings, converted to moment magnitude scale for Mw < 6.0. MW-HARVARD are moment magnitudes as given by the Harvard CMT catalogue on the web site of the Harvard University. Date and origin time 1976/10/01 1977/03/04 1979/05/31 1979/09/11 1981/07/18 1985/08/01 1986/08/30 1990/05/30 1990/05/31 1995/09/06 1995/09/19 1997/11/18 1998/03/13 1999/04/04 1999/04/28 2000/04/06 17:50:43.2 19:21:54.1 07:20:06.3 15:36:54.2 00:02:58.6 14:35:04.3 21:28:35.7 10:40:06.4 00:17:47.9 10:58:45.9 19:32:21.4 11:23:16.3 13:14:37.2 01:21:12.6 08:47:56.9 00:10:39.9 Latitude (deg. N) 45.68 45.77 45.55 45.56 45.69 45.73 45.52 45.83 45.85 45.53 45.63 45.78 45.51 45.70 45.47 45.69 Longitude (deg. E) 26.49 26.76 26.33 26.30 26.42 26.62 26.49 26.89 26.91 26.39 26.58 26.72 26.33 26.44 26.32 26.68 Depth (km) 146 94 120 154 166 93 131 91 87 120 143 126 165 146 145 136 Mw 6.0 7.4 5.3 5.3 5.5 5.8 7.1 6.9 6.4 4.1 4.1 4.6 4.7 3.7 5.3 5.0 Mw-Harvard 7.5 5.2 5.1 5.2 7.2 6.9 6.3 5.2 5.4 5.1 Table 2. The source parameters (corner frequency, high frequency spectral slope, pulse duration and moment magnitude) from P- and Swaves recorded at the stations listed in the last column. The standard deviation and the number of observations (in parenthesis) are also given, when at least two determinations are available. Event date & time 1976/10/01 17:50 1977/03/04 19:21 1979/05/31 07:20 1979/09/11 15:36 1981/07/18 00:02 1985/08/01 14:35 1986/08/30 21:28 1990/05/30 10:40 1990/05/31 00:17 1995/09/06 10:58 1995/09/19 19:32 1997/11/18 11:23 1998/03/13 13 :14 1999/04/04 01:21 1999/04/28 08:47 2000/04/06 00:10 fcP (Hz) 0.35 +0.10 (3) 0.16 +0.06 (3) 0.30 ± 0 . 0 1 (2) 0.4 ±0.06 (2) 0.2 +0.06 (3) 0.19 +0.11 (30) 0.30 ±0.13 (21) 3.66 1.55 1.67 4.03 0.52 ±0.27 (8) 0.84 ±0.52 (7) fcs (Hz) 0.26 ±0.06 (3) 0.1 ±0.02 (3) 0.26 ±0.04 (2) 0.32 +0.03 (2) 0.28 + 0 . 0 1 (2) 0.33 0.13 +0.02 (3) 0.11 ±0.05 (19) 0.14 ±0.07 (16) 2.43 ±0.04 (2) 2.33 ±0.04 (2) 1.36 ± 0 . 2 8 (2) 1.18 ± 0 . 0 6 (2) 2.9 +0.42 (2) 0.57 ± 0 . 5 1 (6) 1.08 + 0 . 9 2 (4) YP Ys TP (S) 1.97 ±0.1 (3) 2.73 ± 0 . 1 5 (3) 2.40 ±0.26 (2) 1.87 ± 0 . 5 8 (2) 3.25 ±0.32 (3) 2.11 ±0.45 (30) 2.05 ±0.43 (21) 1.98 1.57 2.11 1.89 2.21 + 0 . 2 6 (8) 2.54 ±0.44 (7) 1.42 ±0.22 (3) 1.80 ± 0 . 1 5 (3) 1.78 ± 0 . 1 1 (2) 1.90 ±0.08 (2) 1.72 ±0.24 (2) 1.84 2.02 ±0.10 (3) 1.94 +0.22 (19) 1.98 ± 0 . 4 6 (16) 2.13 ±0.14 (2) 1.87 +0.06 (2) 1.92 ±0.12 (2) 1.93 + 0 . 1 (2) 1.72 ± 0 . 0 8 (2) 2.0 ±0.17 (6) 2.16 ±0.30 (4) 2.8 ± 0 . 8 (2) 10.1 ± 1 . 1 (3) 2.6 +0.0 (2) 2.0 + 0 . 1 (2) 4.4 +0.2 (3) 5.1 MWP Mws Recording Station 4.7 ±0.2 (3) 7.0 ±0.2 (3) 4.8 GRF +0.1 (3) 6.9 GRF +0.1 (3) 5.0 GRF +0.1 (2) 4.8 GRF +0.1 (2) 4.9 GRF ±0.1 (2) 5.0 GRF (s) 12.9 ±3.7 (2) 3.5 3.4 ± 0 . 8 (2) 3.5 ± 2 . 1 (2) 5.9 5.1 ±0.0 (2) 4.8 ±0.1 9.2 ±1.1 (3) 9.2 (2) 7.1 ±0.2 (3) 6.7 ±0.2 (27) 6.0 +0.2 (16) ±0.2 (13) 5.9 ± 0 . 1 (6) 4.6 MLR 3.8 +0.1 (2) 4.0 MLR +2.6 (18) 4.1 ±2.6 (10) 8.7 ±1.4 (13) 0.6 ±4.5 (13) 0.5 0.3 + 0 . 1 (2) 0.5 1.0 ±0.0 (2) 0.6 4.7 0.7 ±0.2 (2) 1.1 4.8 ±0.2 (2) 0.3 0.4 2.0 + 0 . 1 (2) 3.9 ±0.7 (5) 1.8 ±0.9 (6) ±2.7 (6) 2.9 ±1.9 (6) 6.8 ±0.1 6.7 5.2 ±0.3 (8) 5.1 ±0.2 (7) AQU, HIA, AQU, HIA, GRF, ARU, CCM, ANMO GRF, ARU, CCM, ANMO ± 0 . 0 (2) 5.0 ± 0 . 1 (2) MLR 5.0 MLR +0.1 4.0 GRF (2) 3.7 ± 0 . 1 (2) 5.1 + 0 . 2 (6) 5.1 +0.2 (4) MLR MLR, T R I , GRF, OBN, ARU, BGCA MLR, T R I , GRF, ARU 21° 24° * 27° ** * - • * * 48° * * 30° 'v| * • * +» 48° •*v i * * "BaiaMare + * • . t' * • ** *<>'^> ttfoa ****V ******* ***:***• i. • 46° ***** *• * * * • * *** \ y in ••••** , * i * s 44° * * *•••••/•KOTtodw * i ••** LEGEND OGEOFON station MLR •*• ** •" . a i~ 44° mWmsmmm * Q Subcrustal cqs. Mw>4 • 21° • MB—1 %***" t» + • • 46° 24° 27° 30° Figure 1. Vrancea seismogenic zone. Intermediate-depth earthquakes with magnitudes Mw > 4.0 from the up-to-date ROMPLUS catalogue are plotted, together with the location of the GEOFON station MLR. Mw - Romplus catalog Figure 3. Moment magnitudes determined in this study versus moment magnitudes of ROMPLUS catalogue. The bisecting line is also plotted. 11 GRF N, 1986/08/30 21:28 - 1E»015 - Frequency (Hz) 0.1 Frequency (Hz) Frequency (Hz) Figure 2. Examples of the considered waveforms and spectra of Vrancea intermediate-depth earthquakes. The signal (P-wave) and noise spectra are plotted together with the chosen spectral model. The example on the left represents a moderate earthquake (Mw 4.7) recorded at the local station MLR, the example at the center of the figure represents the strong 1986 event (Mw 7.1) as recorded by the Grafenberg station, at regional distance. The example on the right represents a moderate earthquake (M w 4.7), which could not be used in this study because of the low signal to noise ratio, as recorded at the Grafenberg station. Stations + D GRF Mw=-2log(fc)+5.13 RMS=0.417 CCM ARU • ANMO O HIA A AQU • MLR • OBN * TRI 0.01 0.1 1 10 Fc - P waves (Hz) Figure 4. M\y (from ROMPLUS catalogue) versus P-wave corner frequency for Vrancea intermediate-depth earthquakes. The dashed line represents the data best fit, and it is defined by the equation given in the upper right-hand corner. The fit RMS (computed as the squared root of the residual mean square) is also given. The solid lines represent the approximate average trends for shallow earthquakes, for 1 and 10 MPA stress drop. 13 Stations + GRF Mw=-2 log(fc)+4.84 ARU RMS=0.298 • ANMO o A • e * HIA AQU MLR BGCA TRI •» \ \ 3.5 0.1 0.01 1 10 Fc - S waves (Hz) Figure 5. Mw (from ROMPLUS catalogue) versus S-wave corner frequency for Vrancea intermediate-depth earthquakes. The dashed line is the data best fit, and it is defined by the equation given in the upper right-hand corner. The fit RMS (computed as the squared root of the residual mean square) is also given. The solid lines represent the approximate average trends for shallow earthquakes, for 1 and 10 MPA stress drop. 14 Stations + D GRF O ARU • ANMO o A Mw=-2 Iog(1/T)+5.02 RMS=0.394 CCM HIA AQU • MLR • OBN * TRI 0.01 Figure 6. Mw (from ROMPLUS catalogue) versus the inverse of the P-wave pulse duration for Vrancea intermediate-depth earthquakes. The dashed line is the data best fit, and it is defined by the equation given in the upper right-hand corner. The fit RMS (computed as the squared root of the residual mean square) is also given. The solid lines represent the approximate average trends for shallow earthquakes, for 1 and 10 MPA stress drop. 15 Stations + GRF • ANMO o A HIA AQU • MLR e * TRI Mw=-2log(1/T)+4.61 ARU BGCA RMS=0.428 6.5 • X \ \ iX • •\ 4.5 \ 0.01 0.1 1 \ 10 1/Tau - S waves (Hz) Figure 7. Mw (from ROMPLUS catalogue) versus the inverse of the S-wave pulse duration for Vrancea intermediate-depth earthquakes. The dashed line is the data best fit, and it is defined by the equation given in the upper right-hand corner. The fit RMS (computed as the squared root of the residual mean square) is also given. The solid lines represent the approximate average trends for shallow earthquakes, for 1 and 10 MPA stress drop. 16 log(i/T) = log(ic) - 0.004 RMS=0.126 Log(fc) - P waves Figure 8. Inverse of the pulse duration versus fc for P-wave of Vrancea intermediatedepth earthquakes. The best fitting line, defined by the equation given in the upper right-hand corner, is plotted. Iog(1/T) = log(!c) - 0 . 0 3 6 BHS=0.100 Log(fc) - S waves Figure 9. Inverse of the pulse duration versus fc for S-wave of Vrancea intermediatedepth earthquakes. The best fitting line, defined by the equation given in the upper right-hand corner, is plotted. 17 Stations + D ++ GRF CCM ARU ANMO O HIA A AQU • MLR A OBN * TRI + • to + r CD > 2.5 to A ±+ 5 CL B • CD it • i * + O • • Mw Figure 10. Slope of the high-frequency decay for P-wave spectra of Vrancea intermediate-depth earthquakes versus Mw (from ROMPLUS catalogue). The stations providing data are identified by different symbols. Stations + GRF • ANMO A AQU • MLR BGCA o e ARU HIA TRI o 2.5 * A 4 + o -^ • • + • 1 1 1 • A 4. + • 1 Mw Figure 11. Slope of the high-frequency decay for S-wave spectra of Vrancea intermediate-depth earthquakes versus Mw (from ROMPLUS catalogue). The stations providing data are identified by different symbols. 18