1. Attenuation of High frequency P and S Waves in Khorasan Province, Iran Jafar Shoja-Taheri, and Mohsen Farrokhi Earthquake Research Center, Ferdowsi University of Mashad, Abstract Hz to 0.0011 at 36 Seismic wave attenuation of direct shear and P waves was investigated for the crust of Khorasan province of Iran, using 749 seismograms recorded by local earthquakes with ML between 1.0 and 5.6 occurred during the period between 2002 and 2006. Single station method (extended coda normalization method) was applied to analyze Q P -1 and Q S -1 by using the ratios of P and S waves to coda-wave amplitude spectra in the frequency range of 1.5 Hz to 36 Hz, as a function of distance. The Q P and Q S are analyzed by using 66 laps times from 22.5 seconds to 87.5 seconds for center-frequencies of 1.5, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 18, 20, 24, 28, 32, and 36 Hz. The estimated values of Q P -1 and Q S -1 are highly frequency-dependent and decreasing respectively from 0.016 and 0.014 at 1.5 Hz to 0.0011 at 36 Hz. They also decrease with the increase of lapse time and earthquake source distance. This indicates that the value of quality factor, Q, increases with depth. The estimated values of Q P -1 and Q S -1 are (0.021 ± 0.002)f (-0.79 ± 0.06) and (0.018 ± 0.001)f (-0.76 ± 0.03) respectively. The relatively low value of Q-factor obtained in this study agrees with the value which is generally expected for a seismically active region such as Khorasan province of Iran. Introduction The Khorasan province in northeastern Iran (the area under study) extends from 30 to 40 degrees north latitude and from 52 to 62 degrees east longitude (Fig. 1). The major tectonic provinces of the region are the Kopeh Dogh folded belt (Tchalenko, 1975), the eastern Alborz and central and eastern Iran province, an area of complex block movement. Near its southwest corner, the region approaches the Zagros main thrust line. Along its eastern edge, the region is bound by the north-south trending Harirud fault which, despite its current aseismic character and pre- Jurassic age (Stöcklin, 1974), serves as a boundary between the aseismic zone of western Afghanistan and the highly seismic region of northeastern Iran (Shoja-Taheri and Niazi, 1981). The Khorasan province, as a region with a high level of seismic activity, has suffered numerous destructive earthquakes throughout its long history including four major ones in 1968, 1978, 1979 and 1997. Attenuation of seismic waves due to intrinsic absorption and scattering is expressed by the inverse of the quality factor, Q -1. The value of Q -1 is a very important parameter used to infer the material and physical conditions of the earth’s interior. In particular, the knowledge of the Q -1 value is indispensable for the quantitative prediction of strong ground motion from the viewpoint of earthquake-resistant design of construction. Hence numerous studies of Q -1 have been undertaken worldwide and concentrated mainly on seismically active zones and/or densely populated industrial areas. These studies show relatively higher Q -1 in seismically active areas than in stable areas. It is now generally accepted that observed Q increases with frequencies between 0.1 and 30 Hz. This frequency dependence, which usually is expressed by the relation Q(f) = Q o ( m), is often found to be stronger with increasing tectonic activity and is often explained as being related to the decrease of homogeneity in. Another common observation is that Q seems to increase with depth which possibly could be a result of increasing pressure and rate of homogeneity. Rovelli (1984) pointed out that increase in Q c with depth will cause Q c to be dependent on the length of the time window chosen for analysis, since waves arriving later in the seismogram may be scattered from deeper parts of crust than energy arriving earlier. For northeastern Iran several reports about Q are available. The Q (1 Hz) value of about 160 was recently reported by Shoja-Taheri, et al. (2008) using the data recorded in the range of regional distances in northeastern. This low-Q value agrees with the results given by Nuttli (1980). He showed that the attenuation of 1- sec period crustal phases in Iran is relatively high with an apparent Q of 200 for L g, 150 for S n, and 125 for P g. Sarker and Abers (1998) obtained Q S = 309 f 0.30 and Q c = 195 f 0.65 for Kopet Dagh in the north of Khorasan province. Aki and Chouet (1975) found a model which could explain the existence and decay of Coda waves for source and scatterer in a zero distance. Using the similarities between characteristics of coda waves and S waves, Aki (1980) proposed a single station method for measuring attenuation by normalizing direct S wave’s amplitude by S-coda’s amplitude. Employment of this method would eliminate the influence of source and site effect so that attenuation can be studied in more details. Yoshimoto et al. (1993) extended the conventional coda normalization method to measure Q p -1 based on the similarity of source spectral amplitude of P and S waves. The purpose of this article is to estimate the frequency-dependent attenuation of P and S waves in the range of local distances in the Khorasan province, northeastern Iran. Figure 1. The station location (triangle and station ID) and the path (thin gray line) from the epicenter (open circle) to station. Data We used waveforms recorded by three-component, high-dynamic-range seismometers. The data come from six permanent stations of the Khorasan Seismic Network (KHSN) equipped with Guralp CMG-3T (flat response between 0.01 and 50 Hz) velocimeters, connected through dial- up lines to the acquisition center. Data acquisition was performed at a sampling rate of 125 samples per second. Figure 1 shows the KHSN stations and epicenters of the earthquakes used in this study. The data comprise 749 three-component velocity time series recorded for earthquakes that occurred between 2002 and 2006 with M L magnitudes between 1.0 and 5.6. Table 1 lists the distribution of the number of events versus magnitudes. For evaluation of Q s -1 the earthquakes with hypocentral distances of less than 120 Km and for Q p -1 the earthquakes with hypocentral distances between 40 and 160 km were chosen. Duration of 5 seconds of time windows was selected for P, S, Coda waves, and background noises. P and S wave time windows begin from the onsets of P and S waves respectively. Spectral amplitudes of P and S waves at different frequency bands were determined respectively from the vertical and horizontal N-S and E-W components. Signal to noise ratios for P and S waves were larger than 4 and for Coda waves the ratios were larger than 2. Amplitude of the Coda waves were determined by taking the root mean squares of the amplitudes within each window centered at each lapse time, t c. To estimate the mean values of Q p -1 and Q -1 s and their variation at each frequency band we have chosen 66 5-seconds time windows along each seismograms moving sequentially from 22.5 to 87.5 seconds from the S- wave arrival times (see Figure 2). Table 1 ML ML No.of Eqs. 1.0-1.9 2.0-2.9 3.0-3.9 4.0-4.9 5.0-5.6 20 269 365 85 10 Figure 2. The coda windows of 5- second durations have been chosen from 22.5 seconds (more than twice the S wave travel time, 2t S ) after the S wave arrival times and sequentially move for 66 seconds along the seismograms. Methodology The method was designed to normalize the spectral amplitude of P and S waves by amplitude of coda waves at a constant lapse time. By analyses of coda waves, Aki (1969) concluded that for earthquake with source distances less than 200 Km and for lapse times greater than twice the S wave travel time, the spectral amplitude of coda waves, A C ij, at lapse time t c is independent of hypocentral distance. Aki(1969), Aki and Chouet (1975), Sato (1977), and Aki(1992) concluded that coda waves are the results of scattered waves from heterogeneities in the crust and that the source effects on both coda waves and S waves are similar. To calculate the quality factor of S waves, Aki (1980) showed that one can normalize the spectral amplitude of S waves by the spectral amplitudes of coda waves using Equation 1: Where A ij Direct-S is spectral amplitude of S wave from i th earthquake recorded by j th station, f is frequency, Q s is quality factor of S waves, β is average velocity of S waves, and r ij is hypocentral distance. Using a data set from wide range of azimuths, the effects of radiation pattern can be neglected. From linear regression of the left hand of this equation versus hypocentral distance, Q s -1 can be evaluated. As mentioned, the similarity of the source effects on both S waves and coda waves, the use normalization method would eliminate the source effects and site amplification. Yoshimoto et al. (1993) concluded that the coda normalization method can also be used for evaluation of Q p -1 (Equation 2) since for earthquake with magnitudes less than five the source spectral amplitudes of P and S waves are similar: (1) (2) Where A ij Direct P is the spectral amplitude of P waves, Q p is the quality factor of P waves, and α is the average velocity of P waves. Center frequency (Hz) Frequency band 1.50.5 2-121.0 14-202.0 24-364.0 Table 2 Measurment of Q p -1 and Q -1 Following the base-line correction and a 10% cosine tapering at both ends at each record, we applied 20 different zero phase-shift band-pass Butterworth filters with four-poles to all of the waveform windows of P, S, and the Coda wave windows at each lapse time. Center frequency and band width of the filters are listed in Table 2. We then used the amplitude of the filtered windows for estimating Q p -1 and Q S -1 values. P wave amplitudes were measured using P wave windows from vertical components and the corresponding Coda wave amplitudes were measured from N-S and E-W components. Likewise, S wave and their corresponding Coda wave amplitudes were measured from horizontal components. We then employed the Equations 1 and 2 to make geometrical corrections for the direct waves and also to normalize the P and S wave amplitudes with respect to the Coda wave amplitudes in each frequency band and in each lapse time window. Linear regression of Equations 1, and 2 give the estimate of Q p -1 and Q s -1 values for the area under the study. Table 3 lists the results of the regression for estimated values of Q p -1 and Q s -1 for the recording stations and the average value for Khorasan province. The value of quality factors in each frequency band generally show increase with increasing the lapse time, because the waves arriving later in the seismogram may be scattered from deeper parts of crust than energy arriving earlier. Thus, to obtain unbiased estimate of the quality factors we used 66 lapse times for evaluating means and variations of Q p -1 and Q S -1 values for each station. Figures 4 and 5 illustrate the values of Q p -1 and Q s -1 versus frequency for the recording stations in Khorasan province, northeastern Iran. Q p -1 and Q s -1 are both strongly dependent on frequency in Khorasan province. Their values decrease from 0.015 and 0.013 at 1.5 Hz to 0.0012 at 36 Hz. The Q P -1 /Q s -1 ratio of 1.17f 0.03 is weakly dependent on frequency and its value is larger than unity in the range of frequency band used in this study. In the average, the frequency dependent quality factors in the region are: Q S =56f 0.76 and Q p =48f 0.79. Table 4 lists the Q S and Q p values at 1 Hz and 10 Hz for the recording stations. The low-Q values obtained in this study expectedly represent the high attenuation of seismic waves in the region of Khorasan province with high tectonic activity. The results are generally in good agreement with the previous studies performed in Northeastern Iran (e.g. ; Nuttli, 1980; Shoja-Taheri and Anderson, 1988; Xie, 1993; Sarker and Abers, 1998; Shoja-Taheri, et al., 2008). StationQ s -1 Q p -1 Bojnord( 0.015±0.000)f (-0.68±0.02) (0.020±0.001) f (-0.65±0.04) Birjand( 0.026±0.001)f (-0.83±0.03) ( 0.030±0.002) f (-0.85±0.07) Najafi( 0.015±0.001)f (-1.00±0.05) ( 0.011±0.001) f (-0.64±0.04) Quchan( 0.011±0.000)f (-0.48±0.02) ( 0.020±0.003) f (-0.65±0.09) Sabzevar( 0.022±0.001)f (-0.71±0.02) ( 0.024±0.002) f (-0.80±0.06) Average( 0.018±0.001)f (-0.76±0.02) ( 0.021±0.002) f (-0.79±0.02) Table 3 Figures 4. The Q p -1 values evaluated versus frequency for the recording stations in Khorasan province. Figures 5. The Q S -1 values evaluated versus frequency for the recording stations in Khorasan province. References Aki, K. (1969). Analysis of the seismic coda of local earthquakes as scattered waves, J. Geophys. Res., 74, 615 – 631. Aki, K., and B. Chouet (1975). Origin of coda waves : Source, Attenuation, and scattering effects, J. Geophys. Res. 80, 3322 – 3342. Aki, K. (1980). Attenuation of shear waves in the lithosphere for frequencies from.05 to 25 Hz, Phys. Earth Planet. Inter. 21, 50 – 60. Aki, K. (1992). Scattering conversions P to S versus S to P, Bull. Seismol. Soc. Am. 82, 1969 – 1972. Nuttli, O.W. (1980). The excitation and attenuation of seismic crustal phases in Iran, Bull. Seismol. Soc. Am. 70, 469–485. Rovelli, A. (1984). Seismic Q for the lithosphere of the Montenegro region (Yugoslavia): frequency, depth and time windowing effects, Phys. Earth Planet. Inter. 34, 159-172. Sarker, G., and G. A. Abers (1998). Comparison of seismic body wave and coda wave measures of Q, Pure Appl. Geophys. 153, 665–683. Sato, H. (1977). Energy Propagation including scattering effects : Single isotropic scattering approximation, J. phys. Earth. 25, 27 -41. Shoja-Taheri, J., and M. Niazi (1981). Seismicity of the Iran plateau and bordering regions, Bull. Seismol. Soc. Am. 71, 477–489. Shoja-Taheri, J., S. Naserieh, and H. Ghograni (2007). M L and M W scales in the Iranian Plateau based on the strong-motion records, Bull. Seismol. Soc. Am. 97, 661–669. StationQ s (1Hz)Q s (10Hz)Q p (1Hz)Q p (10Hz) Bojnord6732050223 Birjand3825633233 Najafi6767091397 Quchan9127550223 Sabzevar4523042265 Average6235053268 Table 4 Conclusions Seismic wave attenuation of direct shear and P waves was investigated for the crust of Khorasan province of Iran. Single station method (extended Coda normalization method) was applied to analyze Q p -1 and Q s -1 by using the ratios of P and S waves to Coda-wave amplitude spectra in the frequency range of 1.5 Hz to 36 Hz, as a function of distance. The estimated values of Q p -1 and Q s -1 are highly frequency-dependent and are decreasing respectively from 0.015 and 0.013 at 1.5 Hz to 0.0012 at 36 Hz. The estimated values of Q p -1 and Q s -1 are (0.021 ± 0.002)f (-0.79 ± 0.06) and (0.018 ± 0.001)f (-0.76 ± 0.03) respectively. The ratio of Q P -1 /Q s -1 = 1.17f 0.03 is weakly dependent on frequency and its value is larger than unity in the range of frequency band used in this study. In the average, the frequency dependent quality factors in the region are: Q S =56f 0.76 and Q p =48f 0.79. The low-Q values obtained in this study expectedly represent the high attenuation of seismic waves in the region of Khorasan province with high tectonic activity. The results are generally in good agreement with the previous studies performed in Khorasan province, northeastern Iran.