22
ORIGINAL ARTICLE Site response and source spectra of S waves in the Zagros region, Iran H. Zafarani & B. Hassani Received: 27 February 2012 / Accepted: 4 November 2012 / Published online: 20 November 2012 # Springer Science+Business Media Dordrecht 2012 Abstract S wave amplitude spectra from shallow earthquakes with magnitudes ranging between 4.2 and 6.2 in the Zagros region of Iran that occurred between 1998 and 2008 are used to examine source parameters and site response of S waves. A general- ized inversion scheme has been used to separate the source, propagation path, and local site effects from S wave spectra. For removing the trade-off between source and site terms and propagation effects (includ- ing geometric and anelastic attenuation), the spectral amplitudes of the records used were corrected for attenuation and geometrical spreading function using a path model proposed by Zafarani and Soghrat (Bull Seism Soc Am 102:20312045, 2012) for the region. We assume a Brunes point source model to retrieve source parameters like corner frequency, moment magnitude, and high-frequency fall off coefficient, for each event. When the source spectra are interpreted in terms of Brunes model, the average stress drops obtained are about 7.1 and 5.9 MPa (71 and 59 bars), respectively for the eastern and western Zagros regions. Stress drops range from 1.4 to 35.0 MPa (14 to 350 bars), with no clear dependence on magnitude. The results in terms of stress drop and S wave seismic energy indicate that the Zagros events are more similar to interplate earthquakes of western North America than to intraplate events of eastern North America. The method also pro- vides us with site responses for all 40 stations individ- ually and is an interesting alternative to other methods, such as the H/V method. A new empirical relationship between body-wave magnitudes and moment magni- tude has been proposed for the Iranian plateau using derived seismic moment from the inversion. Keywords Brunes model . Stress drop . Site response . Interplate earthquake . Zagros 1 Introduction The factors influencing earthquake ground motions have been separated into source, path, and site effects. This separation has proven to be useful for under- standing the physical process of strong-motion gener- ation and predicting future seismic motions (see, e.g., Boore 2003). An accurate and reliable quantitative prediction of earthquake strong motions cannot be achieved without understanding the source, site, and path effects. Obtaining source parameters (e.g., stress drop and corner frequency) is challenging because J Seismol (2013) 17:645666 DOI 10.1007/s10950-012-9344-1 H. Zafarani (*) International Institute of Earthquake Engineering and Seismology (IIEES), No. 26, Arghavan St., North Dibajee, Farmanieh, PO Box 19395/3913, Tehran, Iran e-mail: [email protected] B. Hassani Department of Earth Sciences, Western University, Biology & Geological Sciences Bldg., London, Ontario, Canada N6A 5B7 e-mail: [email protected]

Zafarani Hassani 2013 Site Amplification Earthquakes Zagros

Embed Size (px)

Citation preview

  • ORIGINAL ARTICLE

    Site response and source spectra of S waves in the Zagrosregion, Iran

    H. Zafarani & B. Hassani

    Received: 27 February 2012 /Accepted: 4 November 2012 /Published online: 20 November 2012# Springer Science+Business Media Dordrecht 2012

    Abstract S wave amplitude spectra from shallowearthquakes with magnitudes ranging between 4.2and 6.2 in the Zagros region of Iran that occurredbetween 1998 and 2008 are used to examine sourceparameters and site response of S waves. A general-ized inversion scheme has been used to separate thesource, propagation path, and local site effects from Swave spectra. For removing the trade-off betweensource and site terms and propagation effects (includ-ing geometric and anelastic attenuation), the spectralamplitudes of the records used were corrected forattenuation and geometrical spreading function usinga path model proposed by Zafarani and Soghrat (BullSeism Soc Am 102:20312045, 2012) for the region.We assume a Brunes point source model to retrievesource parameters like corner frequency, momentmagnitude, and high-frequency fall off coefficient,for each event. When the source spectra are interpretedin terms of Brunes model, the average stress drops

    obtained are about 7.1 and 5.9 MPa (71 and 59 bars),respectively for the eastern and western Zagros regions.Stress drops range from1.4 to 35.0MPa (14 to 350 bars),with no clear dependence on magnitude. The results interms of stress drop and S wave seismic energy indicatethat the Zagros events are more similar to interplateearthquakes of western North America than to intraplateevents of eastern North America. The method also pro-vides us with site responses for all 40 stations individ-ually and is an interesting alternative to other methods,such as the H/V method. A new empirical relationshipbetween body-wave magnitudes and moment magni-tude has been proposed for the Iranian plateau usingderived seismic moment from the inversion.

    Keywords Brunesmodel . Stressdrop . Site response .

    Interplate earthquake . Zagros

    1 Introduction

    The factors influencing earthquake ground motionshave been separated into source, path, and site effects.This separation has proven to be useful for under-standing the physical process of strong-motion gener-ation and predicting future seismic motions (see, e.g.,Boore 2003). An accurate and reliable quantitativeprediction of earthquake strong motions cannot beachieved without understanding the source, site, andpath effects. Obtaining source parameters (e.g., stressdrop and corner frequency) is challenging because

    J Seismol (2013) 17:645666DOI 10.1007/s10950-012-9344-1

    H. Zafarani (*)International Institute of Earthquake Engineeringand Seismology (IIEES), No. 26, Arghavan St.,North Dibajee, Farmanieh,PO Box 19395/3913, Tehran, Irane-mail: [email protected]

    B. HassaniDepartment of Earth Sciences, Western University,Biology & Geological Sciences Bldg.,London, Ontario, Canada N6A 5B7e-mail: [email protected]

  • they require analysis of the higher-frequency parts ofthe spectrum, where attenuation, scattering, and otherpath effects have a significant effect. Also, groundmotions in the frequency range of 0.515 Hz, whichare of most engineering interest, are strongly affectedby local site conditions, such as complex surface ge-ology and irregular topography. Therefore, the practi-cal way of estimating site effect and source parametersin this frequency range is an empirical approach (Shojiand Kamiyama 2002).

    In recent years, ground-motion prediction equationsfor the Iranian plateau have been successfully devel-oped (Zafarani et al. 2008; Soghrat et al. 2012;Zafarani and Soghrat 2012) based on the stochasticmodeling approach of Boore (2003), but yet there arelimited number of studies which have focused on theregion-specific seismic source and wave propagationparameters of major seismotectonic provinces of Iran(e.g., Nowroozi 2010; Zafarani et al. 2008, 2009,2012; Hassani et al. 2011; Mousavi et al. 2007;Motazedian 2006). Here, the Zagros region, one ofthe most seismically active regions in Iran has beenselected for investigating the source and site terms.The Zagros Mountains in southwestern Iran are alinear intracontinental seismically active fold-and-thrust belt, accommodating nearly one third of thepresent-day convergence rate between Arabia andEurasia (Berberian 1995; Vernant et al. 2004). Thebelt extends for about 1,800 km from northern Iraq,through south-western Iran, to the Strait of Hormuz(Fig. 1). Seismic activity is widely distributed in theregion (see Fig. 1), with many destructive earthquakes.However, fault rapture at depth usually does not prop-agate to the Earths surface in the Zagros, apparentlydue to the presence of a thick salt layer above thebasement, preventing ruptures reaching the surface(e.g., Berberian 1995). The Zagros range can be sub-divided into two main subregions that are distinct intheir topography, geomorphology, exposed stratigra-phy, and seismicity. The elevation of the mountains inthe 100-km-wide north-eastern zone, called the HighZagros (Fig. 1) range from 1,500 to 2,000 m, thetallest mountains reaching up to an elevation of over4,000 m and exposing stratigraphic levels in theMesozoic and Palaeozoic. The 100- to 200-km-widesouth-western zone, called the Simply Folded Belt(SFB), rises from sea level in the SW to 1.5 km inthe NE (Fig. 1) and exposes Palaeozoic strata onlyrarely (except for the Hormuz salt plugs). The SFB is

    dominated by the large, open, linear folds for whichthe range is famous (e.g., Berberian 1995; Ramsey etal. 2008). The southern margin of the High Zagros liesalong the High Zagros fault (Figs. 2 and 3), which isusually the southern limit of exposure of Palaeozoicrocks (Berberian 1995). Most of the present-day de-formation determined by GPS (e.g., Vernant et al.2004) and seismicity occurs in the SFB (Fig. 1), inwhich shortening within young sedimentary rocks isaccommodated as folding along anticlines (e.g.,Berberian 1995; Ramsey et al. 2008). The Zagrosregion frequently experiences small to moderate earth-quakes. However, though the historical catalog of theregion contains events as large as M7.4 (the Silakhorearthquake of 1909; see Ambraseys and Melville1982), there are few strong-motion records of earth-quakes with a magnitude greater than M6 that wouldallow to take an empirical approach for determinationof ground-motion characteristics. The largest earth-quakes (shown in Fig. 1) occurred in the region duringthe instrumental era (i.e., the M7.4 Silakhor earth-quake of 22 January 1909, the M6.7 Ghir earthquakeof 10 April 1972, and the M 6.7 Khorgu earthquake of

    Fig. 1 Seismicity of the Zagros region (M>5.0), 19642010.Major active faults are also shown with solid lines. The boxedregions are the areas shown in Figs. 2 and 3. Hexagram in box 1denotes the location of the Silakhor earthquake of 22 January1909 (M=7.4). The Ghir earthquake of 10 April 1972 (M=6.7),and the Khorgu earthquake of 21 March 1977 (M=6.7), are alsoshown by two pentagrams in box 2. HZ High Zagros zone, SFBsimple fold belt

    646 J Seismol (2013) 17:645666

  • 21 March 1977). Unfortunately, there is no strong-motion record from these three large events.

    Therefore, the appropriate method of predictingground motions for future earthquakes in the Zagros

    47oE 30 48oE 30 49oE 3033oN

    20

    40

    34oN

    20

    40

    35oN

    Asr

    ArnBtn

    Bod

    CCnDAa

    Drd

    Ljb

    NAd

    Sld

    Snr

    Ann

    CaiDsh

    Dhr

    Km1

    Shh

    TAd

    Markazi

    Hamedan

    Kermanshah

    Lorestan54~5

    M

    1

    HighZagros F.

    Sahneh F.

    Main ZagrosReverse F.

    Nahavand F.

    Fig. 2 Epicenter location ofthe earthquakes analyzed forthe present study (circles),strong-motion stations (tri-angles), and paths (solidline) in subregion 1 (areaoutlined in solid box inFig. 1). Major active faultsare also shown with solidlines

    51oE 52oE 53oE 54oE 55oE 56oE26oN

    27oN

    28oN

    29oN

    30oNGom

    HAdJhm

    Krn

    Zdr

    Bah

    Kes

    KenRci

    Rgn

    Persian Gulf

    Be1Be2

    Ber

    Dbn

    Gmh

    Jum

    Ken

    Ner

    Qsm

    SuaTalTmn

    5

    M

    4~5

    Fars

    Kerman

    Hormozgan

    2

    Lar F.

    Beriz F.

    Sarvestan F.

    Bakhtegan F.

    Zagros MountainFront F.

    MFF.

    MFF.

    SabzPushan F.

    HZF.

    HZF.

    Zagrosforedeep F.

    MountainFrontal F.

    Karebas F.

    ZFF.

    Borazjan F.

    Qir F.

    HighZagros F.

    Main ZagrosReverse F.

    Fig. 3 Epicenter location ofthe earthquakes analyzed forthe present study (circles),strong-motion stations (tri-angles), and paths (solidline) in subregion 2 (areaoutlined in solid box inFig. 1). Major active faultsare also shown with solidlines. MFF mountain frontfault, ZFF Zagros foredeepfault, HZF High Zagros fault

    J Seismol (2013) 17:645666 647

  • region should be based on an assumed seismologicalmodel of source and propagation processes, takinginto account that physical models calibrated to the dataof limited magnitude and distance ranges can be usedwith reasonable confidence in predicting motions be-yond those distance and magnitude ranges. To theextent of our knowledge, this study is the first effortto address this problem with strong-motion data forthe Zagros region. In this paper, the generalized inver-sion (GI) of the S wave amplitude spectra from thestrong-motion network data in this region is used toestimate source and site effects. This technique wasfirst presented by Andrews (1986) by recasting themethod of spectral ratios into a generalized inverseproblem and simultaneously solving of the data ofmultiple-recorded events. Since after, the approachhas been used and developed by numerous authors(e.g., Iwata and Irikura 1988; Castro et al. 1990;Boatwright et al. 1991; Hartzell 1992; Harmsen1997; Dutta et al. 2001, 2003; Salazar et al. 2007;Drouet et al. 2008; Tramelli et al. 2010). Here, a largenumber of significant earthquakes in the region wereanalyzed, considering the events that occurred from1998 to the end of 2008. Furthermore, this is the firstsystematic study on the determination of momentmagnitude from strong-motion records for the region.

    2 Data

    We collected three-component waveforms recordedat 40 stations of the Iranian strong-motion network,which have been installed and maintained by theBuilding and House Research Center (BHRC;Mirzaei Alavijeh et al. 1998, 2007). All of the sta-tions are equipped with a digital accelerometer(Kinemetrics model SSA-2). We analyzed 148waveforms from 35 events which occurred fromAugust 1998 to December 2008 and ranged in mag-nitude from 4.2 to 6.2 with a focal depth of less than30 km. The epicentral distances ranged from 10 to100 km. The list of earthquakes used in this studyis given in Appendix A, along with the event num-ber, seismic characteristics, and recording stationsfor each event. Also, the name, geographic coordi-nates, and a three-letter abbreviation, for strong-motion stations is provided in Appendix B. Theregion under study consists of two subregions asdepicted in Fig. 1. As it is clear, subregion 1

    completely belongs to the High Zagros and subre-gion 2 is completely located in the SFB. Figures 2and 3 show a distribution of stations and events usedhere, and also a visual concept of different paths andlinked records is provided using line between sta-tions and corresponding earthquakes. To obtain thefinal database, following steps were performed: (1)earthquakes with reported seismic characteristics,since 1998 installation up to the end of 2008, wereselected in the Zagros region of Iran. (2) Three-component records with only one correctable hori-zontal component were excluded from the database.(3) Earthquakes with at least three, stations with atleast two records, which are linked to the referencesites, compose the final database. As mentioned insimilar studies (e.g., Harmsen 1997), a key factor toachieve reliable results using the GI method is theavailability of adequate number of recorded groundmotions for each event which are distributed invarious directions in order to reduce the directivityeffects and also sufficient number of recorded eventsfor each site. Both of the abovementioned con-straints were applied to the selected dataset. Themagnitude-distance distribution of the dataset usedfor this study is shown in Fig. 4. For the sake ofcompleteness, it should be noted that here, we try todetermine the source spectra of each event separate-ly (see Table 1 below), and therefore the lack orscarcity of data at some magnitude-distance bins isnot a serious/crucial obstacle. We should keep inmind that the path effect (Q factor) has been adoptedfrom an independent study (Zafarani and Soghrat2012). In other words, since we try to investigatesource spectrum of each event individually, we donot need to have a rich catalog in all magnitudes andsince we have taken the Q factor from a separatestudy, we believe that the scarcity/lack of data insome distances is not a great deal. For the sake ofcompleteness, it should be noted that the GI formu-lation is based on the linear assumption and siteamplification (site effects) have been assumed tobe independent of strong-motion level.

    The uncorrected acceleration time series recorded bya given station were corrected for the instrument re-sponse and baseline, following a standard algorithm(Trifunac and Lee 1973). Multiresolution wavelet anal-ysis (Ansari et al. 2010) has been conducted in order toremove undesirable noise from the recorded signals.The data were processed to obtain the Fourier spectra

    648 J Seismol (2013) 17:645666

  • of the S wave windows which contain the strongest partof the shaking (Soghrat et al. 2012). We first determinedthe onset times of S waves on the records using theHusid plot of energy buildup (Husid 1967). Also, toestimate the time of the last arrival of the direct S wave,cumulative root mean square function (McCann andShah 1979) has been used. We then extracted the Swave trains with a cosine-tapered window applied atthe beginning and the end of each window. The lengthof each of these tapers will be 5 % of the total tracelength. The spectrum was smoothed using a 5-pointmoving average filter and interpolated at 20 points,which are equally spaced on the logarithmic scale, be-tween 0.4 and 15 Hz. The Fourier amplitude spectrum(FAS) of horizontal ground motion was computed as theresultant geometric mean of the orthogonal componentsof motion. Figure 5 shows an example of the algorithmused here for determination of S wave window (seeSoghrat et al. 2012 for more details).

    3 Method of analysis

    We use the GI technique of Andrews (1986),implementing the singular value decomposition(SVD) algorithm. Here, we provide a methodoutline. For further details and a complete de-scription of the method, we refer the reader tothe works of Andrews (1986), Harzell (1992),Salazar et al. (2007), and Hassani et al. (2011).Suppose the database contains J stations overwhich I earthquakes have been recorded, thenfollowing Andrews (1986), the shear-wave

    spectrum of the ith event recorded at the jth siteYij () can be written in the frequency domain asa product of a source term Ei(), a path termPij(,R), and a site-effect term Gj():

    Yijf Eif Pij f ;R Gjf 1This factorization implicitly accepts the validity of

    the principle of superposition, which in turn makes itpossible to treat each of these terms as an independentfilter (Salazar et al. 2007). This general representationof the problem also assumes that directional effects inthe source are averaged out by observations at differ-ent azimuths (Hartzell 1992). Here, the path term isrepresented by geometrical spreading (Z(Rij)) and an-elastic attenuation (Q()):

    Pij f ;R Z Rij

    : expp fR ijQsf :bs

    2

    where Rij is the hypocentral distance between theith event and the jth station, Qs() and s are Swave frequency-dependent quality factor and ve-locity, respectively. Because of the trade-off be-tween path attenuation and the high-frequencyfalloff of the spectrum, no attempt was made inthe current study to independently determine Q()(Hartzell 1992). Instead, a propagation term (in-cluding geometric and anelastic attenuation), hasbeen adopted from Zafarani and Soghrat (2012),which is represented as follow:

    ZRij / Rij1 ;Rij < 40 km

    Rij0:5 Rij 40 km

    3

    101 1023

    3.5

    4

    4.5

    5

    5.5

    6

    6.5

    7

    7.5

    Mw

    Distance (Rhypo, Km)

    Eastern ZagrosWestern Zagros

    Fig. 4 Magnitude-distancedistribution of records usedin this study

    J Seismol (2013) 17:645666 649

  • Qf 153f 0:83 4A site effect Gj() representing combination of site

    amplification and the diminution function (path-inde-pendent loss of energy) terms was also considered as:

    Gj f Aj f exp pk0f 5

    where Aj()is the jth station amplification function and0 represents the diminution parameter or zero-distance kappa factor (Anderson and Hough 1984).

    We first determined for all accelerometric stationsthrough the conventional technique based on the esti-mate of the spectral decay of the Fourier amplitudespectra at higher frequencies, generally more than

    Table 1 List of source parametersseismic moment (Mo), moment magnitude (Mw), stress drop (), corner frequency ( fc), radius (r),and radiated energy (Es) estimated from the inversion of S wave spectra

    Event No. Mo (Nm) Mw (MPa) fc (Hz) Es (J) r (km)

    1 2.51E+16 4.9 3.2 0.86 2 2.52E+11 1.5

    2 1.58E+18 6.1 4.5 0.24 2 1.10E+13 5.4

    3 2.82E+17 5.6 1.4 0.29 2 8.07E+11 4.5

    4 4.31E+15 4.4 4.4 1.73 2 7.94E+10 0.8

    5 1.21E+16 4.7 7.1 1.43 2 2.52E+11 0.9

    6 4.80E+15 4.4 6.5 1.90 2 9.66E+10 0.7

    7 7.63E+15 4.6 12 1.99 2 2.78E+11 0.7

    8 5.01E+16 5.1 4 0.74 2 6.15E+11 1.8

    9 1.52E+16 4.8 22.4 1.95 2 9.23E+11 0.7

    10 4.52E+15 4.4 6.4 1.93 2 1.14E+11 0.7

    11 2.00E+17 5.5 2.4 0.39 2 1.38E+12 3.3

    12 2.13E+16 4.9 1.7 0.74 2 1.17E+11 1.8

    13 8.50E+15 4.6 4.1 1.35 2 1.34E+11 1.0

    14 1.32E+16 4.7 6.8 1.37 2 2.63E+11 0.9

    15 5.01E+16 5.1 1.7 0.56 2 2.52E+11 2.3

    16 3.65E+15 4.3 2.6 1.53 2 3.86E+10 0.9

    17 7.76E+17 5.9 13.4 0.44 2 3.41E+13 2.9

    18 5.62E+17 5.8 12.8 0.49 2 3.83E+13 2.7

    19 5.01E+16 5.1 15.6 1.16 2 3.94E+12 1.1

    20 7.66E+17 5.9 3.7 0.29 2 7.92E+12 4.5

    21 9.24E+15 4.6 15.4 2.03 2 4.34E+11 0.6

    22 4.50E+16 5.1 9.6 1.02 2 1.34E+12 1.3

    23 1.12E+18 6.0 14.5 0.40 2 3.69E+13 3.2

    24 1.19E+16 4.7 6.4 1.40 2 2.72E+11 0.9

    25 3.55E+16 5.0 9.3 1.10 2 1.05E+12 1.2

    26 8.13E+14 3.9 35 6.01 2 6.60E+10 0.2

    27 2.87E+15 4.3 8.9 2.50 2 8.03E+10 0.5

    28 6.26E+17 5.8 3 0.29 2 5.17E+12 4.5

    29 1.00E+17 5.3 11.7 0.84 2 3.73E+12 1.6

    30 2.24E+18 6.2 8.9 0.27 2 6.15E+13 4.8

    31 5.01E+16 5.1 12.5 1.08 2 2.15E+12 1.2

    32 7.00E+16 5.2 8.9 0.86 2 3.09E+12 1.5

    33 2.00E+17 5.5 6.4 0.54 2 4.60E+12 2.4

    34 6.59E+16 5.2 5.4 0.74 2 1.65E+12 1.7

    35 2.63E+16 4.9 6.1 1.05 2 5.71E+11 1.2

    650 J Seismol (2013) 17:645666

  • 5 Hz, in a log-linear plot (Anderson and Hough 1984).An average for each station is obtained from thegeometric mean of the Fourier spectra (see Fig. 6).Then, the zero-distance kappa factor was determinedfrom the best-fitted line (=rR+0) through the dis-tribution of kappa factors as a function of distances(Anderson and Hough 1984). The best-fit regressionline is =0.0002R+0.043, which suggests a value of0.043 for 0. The kappa factor of 0.024 for the verticalcomponent is estimated based on the same procedure(Fig. 7). The high dispersion here, which is also com-mon in almost all kappa determination studies (see,e.g., Motazedian 2006) is due to the fact that such asimple factor could not explain/represent the complexphenomena of wave propagation in the earths crustand near-surface soil-sedimentary layers. Also, thestatistical uncertainty is estimated as the 1 standarddeviation range of the linear regression.

    Substituting Eqs. 2, 4, and 5 into Eq. 1, the model isthen linearized by taking the natural logarithms. If weconsider a total number of recordings (N), correspondingto a number of earthquakes (I) recorded at a total number

    of stations (J), then a compact matrix formulation forEq. 6 can be represented as follows:

    Y A path source1; ::::::i; I

    site1; ::::::j; J

    X 6

    where [Y] represents a vector containing theground-motion amplitude corrected by the geomet-rical spreading and the diminution functions; inaddition, the last seven rows stand for predefinedvalues for seven reference sites (see below), andeach of its member represents a vector with nrows, when n is the number of selected points indesired frequency range; [A] denotes a sparse ma-trix containing only three nonzero elements ineach row (from rows 1 to n). The first columncontains a nonzero vector with n rows while othercolumns represent an nn diagonal matrix. [X]represents a matrix holding I+J+1 terms, namely,the vector solution containing the path term Q f ,source effect In(Ei()) and the site amplificationterm In(Aj()) when each of its members represents

    0 10 20 30 40 50 60 70 80100

    0100

    Event No.30, Station No.37, Comb1x(cm

    /s2)

    0 10 20 30 40 50 60 70 805

    0

    5

    N(cm

    /s2)

    0 10 20 30 40 50 60 70 80100

    0100 t1 t2

    x(cm

    /s2)

    Corr

    e cte

    d

    0 10 20 30 40 50 60 70 800.05

    0.5

    1

    E/E t

    0 10 20 30 40 50 60 70 800

    50

    100

    e(cm

    /s2)

    0 10 20 30 40 50 60 70 800

    10

    20t2

    c(cm/

    s2 )

    0 10 20 30 40 50 60 70 80100

    0

    100

    t(sec)

    Swav

    e

    (cm/s2

    )

    101 100 101 102100

    101

    102Event No.30, Station No.37, Comb1

    f(Hz)

    acce

    lera

    tion

    spec

    trum

    (cm/s)

    t1

    Fig. 5 An example of selecting the direct shear-wave windowfor an event that occurred on 10 Sep. 2008 recorded in the LenjAb station. Symbols t1 and t2 denote the estimated arrival andend times of the direct S wave, respectively. Left (from top tobottom), the uncorrected acceleration record, extracted noise

    identified by the wavelet de-noising method, the wavelet de-noised signal, and its Husid plot, acceleration envelope function,cumulative RMS function, the S wave portion of the correctedsignal; right, the acceleration Fourier amplitude spectrum of Swave window

    J Seismol (2013) 17:645666 651

  • a vector with n rows. Equation 6 is solved for (X)in the frequency range of interest, using a SVD

    algorithm (Andrews 1986) which provides a nu-merically robust solution to the problem.

    0 20 40 60 80 100 120 140 160 180 2000

    0.02

    0.04

    0.06

    0.08

    0.1Horizental Components

    0 20 40 60 80 100 120 140 160 180 200

    0

    0.05

    0.1

    0.02

    R(km)

    Vertical Components

    = (0.00030.00003)R +(0.0240.0016)

    = (0.00020.00003)R +(0.0430.0014)

    Fig. 7 The distribution of kappa factor versus distance for horizontal (top) and vertical (bottom) components. Gray lines denote the 1standard deviation range

    0 10 20 30 40102

    101

    100

    101Event No.10, Station No.8

    FAS(

    cm/s)

    0 10 20 30 40102

    101

    100

    101Event No.5, Station No.23

    FAS(

    cm/s)

    0 10 20 30 40102

    100

    102Event No.9, Station No.28

    0 10 20 30 40104

    102

    100

    102Event No.4, Station No.31

    0 10 20 30 40102

    101

    100

    101Event No.12, Station No.3

    0 10 20 30 40102

    101

    100

    101Event No.14, Station No.4

    0 10 20 30 40104

    102

    100

    102Event No.15, Station No.2

    FAS(

    cm/s)

    Frequency(Hz) 0 10 20 30 40104

    102

    100

    102Event No.22, Station No.16

    Frequency(Hz)0 10 20 30 40

    101

    100

    101

    102Event No.32, Station No.30

    Frequency(Hz)

    =0.046

    =0.040

    =0.051

    =0.032=0.064

    =0.041

    =0.068 =0.058 =0.044

    Fig. 6 Determination of kappa factor for average of horizontal components. The straight lines show the regression fits to logarithm ofspectral amplitude versus frequency

    652 J Seismol (2013) 17:645666

  • Andrews (1986) pointed out the presence of oneundetermined degree of freedom in this system ofequations which can be removed by specifying theamplification at selected reference sites to approx-imately be a constant value or by assuming pre-defined shapes for source spectra (e.g., 2 model).In the current study, the database contains fiveseparate groups of data (see Figs. 2 and 3); there-fore, we need to select at least five reference sites(i.e., one for each group). The inversion analysishas been done for each network/subsets of data(no for the whole set of data). This is due to thefact that the formulation of the GI method (seeEq. 6) needs the connectivity of data for each network.This requirement is fulfilled by considering five separatesubsets of data (see Figs. 2 and 3). The main assumptionabout reference site, is surmising a frequency indepen-dent site response (see Hassani et al. 2011 for moredetails). Here, a preliminary inversion has been doneto investigate relative site responses, assuming that themean of the whole set of stations is free of site effect ateach frequency (Drouet et al. 2008).

    The shear-wave velocity values in the upper-most 30 m (30) values were available at 31strong-motion stations of the studied area. Thesestations with their corresponding shear-wave ve-locity are tabulated in Appendix B. Taking intoaccount the 30 values, surface geological condi-tions, if available, the H/V ratios and the resultingresponses of this inversion, stations with the low-est relative values of response have been consid-ered as reference stations and those showingamplification peaks or troughs compared withthe mean have been removed from the set ofreference stations. Chen and Atkinson (2002),Siddiqqi and Atkinson (2002), and Sokolov etal. (2005) showed that the H/V ratios for rocksites, when averaged over many recordings, arecorrelated with the general geological conditionsand may be used as a simple and useful estimateof an amplification effect. Here, the H/V ratios atthe reference sites with necessary modificationshave been used as a constraint to remove theabovementioned undetermined degree of freedomand a second inversion was performed using thisreference condition. Following Drouet et al.(2008), we finally checked on the final resultsthat the stations used in the reference still havea flat transfer function not too far from a

    theoretical rock-site response (i.e., a flat ampli-tude response in the frequencies of engineeringinterest).

    4 Inversion results

    Applying the inversion method to the selected data-base, the source spectra, site responses, and S wavequality factors were obtained simultaneously. Theresults are presented and discussed in the followingsections.

    4.1 Source model

    The source term Ei() of ith event derived from theinversion can be written as

    Eif C4p2f 2 M 0i 7and

    C RVF=4psbs3R0 8where R=0.55 is the average shear-wave radiation pat-

    tern, F=2 is the free surface amplification, V 1= 2p isintroduced to account for the partition of total shear-waveenergy into two horizontal components, s=2.8 g/cm

    3 ands=3.5 km/s are the mass density and the shear-wavevelocity in the vicinity of the earthquake source, R0=1 km is a reference distance, and M0i() is the momentrate spectrum that can be expressed in the form

    M 0if M0i1 f =fcig i 9

    where M0i(), 0i, and i respectively are the scalarmoment, corner-, and the high-frequency spectral fall-off coefficient associated with the ith earthquake. Themost commonly used model of the earthquake sourcespectrum is the theoretical 2 model introduced byAki (1967) using Brune (1970, 1971) source scaling.Since for small to moderate events, the source issmall, and likely to be simple, it is valid to ignorefinite-source effects and assume a single-corner-frequency point source to model the spectral ampli-tudes of S waves. Here, we assume that the sourcespectra for all events are approximated by an 2

    source model (i.e., i=2). In the second stage ofinversion, we visually estimated the initial corner

    J Seismol (2013) 17:645666 653

  • frequency (0in) from resultant source spectra, andused low- and high-frequency level of the sourcespectra to estimate the initial guess of the scalarmoment (M0in(), consequently MWin). A gridsearch analysis was carried out to estimate theoptimal values of abovementioned parameters foreach event computed from M0in() values of Eq. 9,so that residuals between estimated and observedspectral amplitudes become minimum. The follow-ing ranges of parameters have been considered:MWin 0.1 MWi MWin + 0.1 for events withreported moment magnitude and M(b, L, S) 0.5MM(b, L, S)+0.5 for events without reported mo-ment magnitude, and 0.2
  • 2003, Eq. 12). The value of i in general does notdepend on earthquake size, and this led to the scaling ofearthquakes in terms of source dimensions and averageslip. Aki (1967) has shown that for constant stress dropM0c3 is a constant, and this dependence of the cornerfrequency c on the moment M0 controls the scaling ofthe spectral shapes. Repeating the regression analysiswith the slope fixed at 3, corresponding to similarity inthe earthquake source and constant stress drop, we get

    logM0 16:52 0:056 3 log fc 13The plot of the relation in Eq. 13 is shown by a thick

    solid line in Fig. 10, and the shaded area in the same

    figure shows 1 standard deviation of the mean value.Equation 13 can be rewritten as M0c3=3.311016Nm1s3 corresponding to a constant stress drop of6.6 MPa. The results (comparison of Figs. 7 and8) show that there is no significant difference andstrong deviation from similarity assumption. Thisfact is of utmost important in the point-sourcestochastic simulations which will be performingto predict ground motions from the future largeearthquakes in the region. Hassani et al. (2011)obtained a relation M0c3=2.481016Nm1s3 forsmall to moderate earthquakes in the Central-Eastern region of Iran. Also, Zafarani et al.(2012) have found a relationship M0c3=6.811016Nm1 s3 for events in the Alborz region,Northern Iran. Taking to account the estimatedstress drops of the two subregions (areas 1 and2) separately, Eq. 13 can be rewritten as M0c

    3=2.971016Nm1s3 corresponding to a constantstress drop of 5.9 MPa for the western Zagros (area 1in Fig. 1 belongs to the High Zagros region) andM0c

    3=3.561016Nm1s3 corresponding to a constant stressdrop of 7.1 MPa for the eastern Zagros (area 2 inFig. 1 belongs to the SFB region). The plot ofboth relations are shown by a thick solid line inFig. 11, and the shaded area in the same figureshows 1 standard deviation of the mean value. Asit is clear, there is no significant difference be-tween source characteristics of two the subregions.This was expected because both subregions belongto the Zagros seismotectonic zone, having an inter-plate regime (Zafarani and Soghrat 2012).

    After correction for all known path and siteeffects, including geometrical spreading, attenuation,and site amplifications (including kappa), the radi-ated S wave energy ESi for the ith event wascalculated using the following relation (Vassiliouand Kanamori 1982)

    ESi AZ1

    12pf M oif j j

    2

    df 14

    where A 15pa51 10pb5S1

    and is the Pwave velocity near the source region. The obtainedvalues of ESi may be underestimated to someextent because the frequency band limited of thespectral data was used to evaluate Eq. 14. The

    101 100 101 1021012

    1014

    1016

    1018

    1020Se

    ismic

    mom

    ent(N

    .m)

    fc(Hz)

    Mofc

    2.54 = 3.47x1016 N.m/s3

    Fig. 9 Plot of seismic moment versus corner frequencyobtained from S wave spectral analysis. The regression line(solid) is shown with the 1 standard deviation range (shaded)

    101 100 101 1021012

    1014

    1016

    1018

    1020

    Seism

    ic m

    omen

    t(N.m

    )

    fc(Hz)

    = 3.5 MPa

    = 16.1 Mpa

    Mofc

    3 = 3.31x 10 16 N.m/s3

    = 6.6 MPa

    Fig. 10 Plot of seismic moment versus corner frequencyobtained from S wave spectral analysis by fixing the slope inEq. 12 at 3. The regression line (solid) is shown with the 1standard deviation range (shaded)

    J Seismol (2013) 17:645666 655

  • relation between Es and M0 obtained in this studycan be expressed as

    logEs 4:68 0:061 logM0 15A comparison between this study and similar

    studies is depicted in Fig. 12. The dashed line istaken from Hassani et al. (2011) obtained for theCentral-Eastern region of Iran, and the solid lineis the result of Zafarani et al. (2012) for theAlborz region, Northern Iran. The results showthat earthquakes in the Northern part of Iran havehigher energy and higher stress drops than Zagros

    and Central-Eastern parts of Iran. According toScholz et al. (1986), an earthquake that occurson a well-defined plate boundary such as, say, the SanAndreas fault, is clearly an interplate earthquake,and one that occurs in a midplate region far fromany known plate boundary is clearly intraplate.The Zagros region is situated on the deformednorthern margin of the Arabian continental plate,near the boundary of Arabian and southernEurasian plate (Talebian and Jackson 2004), andtherefore the characteristics of these events aremore like interplate earthquakes of western NorthAmerica (see also Silva et al. 1997). The resultsof this study also show the similarity of theZagros earthquakes with the western USA thaneastern North America from the view point ofsource characteristics like spectral amplitudes andlocal stress drops. In contrary, Northern Iran is farmore away from the plate boundary and its earth-quakes clearly can be described as intraplate/mid-plate events (Zafarani and Soghrat 2012). Already,Shoja-Taheri and Niazi (1981) have shown thatthe seismicity pattern of intraplate subregions ofAlborz and Central-Eastern Iran is diffuse, and thecumulative strain release is about one order ofmagnitude below the two interplate boundariesrepresenting Zagros and Hindukush.

    Finally, using derived seismic moment from thesecond-stage inversion in the current study and theprevious studies by Hassani et al. (2011) and Zafaraniet al. (2012), the body-wave magnitudes (mb), for 46events in the Iranian plateau which were reported by

    101 100 1011012

    1014

    1016

    1018

    1020

    fc(Hz)

    Eastern Zagros

    101 100 1011012

    1014

    1016

    1018

    1020

    Seism

    ic m

    omen

    t(N.m

    )

    fc(Hz)

    Western Zagros

    Mofc

    3 = 2.97x 10 16 N.m/s3

    = 5.9 MPa = 7.1 MpaM

    ofc

    3 = 3.56x 10 16 N.m/s3

    Fig. 11 Plot of seismic mo-ment versus corner frequen-cy obtained from S wavespectral analysis by fixingthe slope in Eq. 12 at 3obtained for the Western(Fig. 2) and Eastern (Fig. 3)Zagros regions. The regres-sion line (solid) is shownwith the 1 standard devia-tion range (shaded)

    1014 1015 1016 1017 1018 1019 1020109

    1010

    1011

    1012

    1013

    1014

    1015

    Ener

    gy(E

    s), N

    .m

    Seismic Moment(M0), N.m

    Es/M

    o = 2.11x 10 5

    (Zagros region)

    Es/M

    o = 8.2x 10 5

    (Northern part of Iran)

    Es/M

    o = 2.5x 10 5

    (CentralEast part of Iran)

    Fig. 12 Plot of seismic energy versus moment. The regressionline (solid) corresponds to a constant stress drop is shown alongwith the 1 standard deviation range (shaded). The result ofZafarani et al. (2012) for the Alborz region, Northern part ofIran is also shown. The broken line is taken from Hassani et al.(2011) obtained for the Central-Eastern part of Iran

    656 J Seismol (2013) 17:645666

  • NEIC, has been used to obtain a relationship betweenmb andMw. The relation can be expressed as follow andis shown in Fig. 13.

    Mw 0:81 0:11mb 0:86 0:50 16

    An empirical global relationship converting mb toMw proposed by Scordilis (2006) has also been shownfor comparison. Using data from Central-Eastern Iran(i.e., 13 events taken from Hassani et al. 2011), leadsto very similar results

    Mw 0:79 0:08mb 0:87 0:35 17

    4.2 Site response

    The next outcome of this study is the site ampli-fication, Aj f for 40 stations which was obtainedin the defined frequency range of 0.4 to 15 Hz.Thick and thin solid lines in Fig. 14 represent themean site amplification factors estimated from theGI at each station and 1 standard deviation, re-spectively. Figure 14 also shows comparisons ofamplification spectra at each site between the GImethod and the earthquake H/V method of Lermoand Chvez-Garca (1993). The method assumesthat the local site conditions do not significantlyinfluence the vertical component of the groundmotion and the H/V spectral ratios may be usedas an indicator of site effects. Here, similar to what

    has been done in Chen and Atkinson (2002) andMotazedian (2006), a combination of amplificationthrough the crustal profile and near-surface atten-uation, kappa, was applied in order to model theH/V ratios. In order to clarify the differences in theestimates of site amplifications, site amplificationfactors for both methods at each station were av-eraged over ten frequency bins with central valuesequally spaced on the logarithmic scale (f=1.13,2.59, 4.05, 5.51, 6.97, 8.43, 9.89, 11.35, 12.81,and 14.27 Hz). The averaged amplification factorsat each frequency bands were plotted for the re-spective methods (Fig. 15). As it is clear, compar-ison of site response values of inversion and thoseof H/V shows that the results are more scattered inhigher frequencies (f>5 Hz). A similar trend canbe found in other studies (e.g., Shojia andKamiyama 2002). Also, comparison of the H/Vresults with the inversion technique reveals thatboth methods practically represent the same shapeof site response.

    Figure 16 shows the obtained amplification val-ues which have been averaged over the entirefrequency band (0.415 Hz) in comparison withthe results of H/V method. For comparison, theaverage amplification over all frequencies obtainedfor Central-Eastern (Hassani et al. 2011) andNorthern Iran regions (Zafarani et al. 2012) arealso shown. In conclusion, it can be said that inan average sense over all frequencies, there is agood agreement in not only dominant period butalso amplification factor between the GI schemeand the H/V method. In order to examine theresult of the inversion method, against the sitecondition, sites were classified in two categoriesas rock and soil. Sites with 30500 m/sare categorized as rock (Zare et al. 1999).The average of site response obtained by inver-sion is shown in Fig. 17, for both categories.A band width of 0.6 Hz was used for the low-frequency band=0.41.0 Hz, and 3.5 Hz forthe high-frequency band = 2.56.0 Hz. Thesolid line shows the best-fitted line between theresults of the current study. Dashed and dottedlines show the similar quantities for East-Central Iran and the Alborz region, NorthernIran, respectively. It can be concluded that site

    3 3.5 4 4.5 5 5.53

    3.5

    4

    4.5

    5

    5.5

    mb

    Mw

    Mw = (0.81 0.11) mb + (0.86 0.50)(This study)

    Mw = (0.85 0.11) mb + (1.03 0.23)3.5 mb 6.5(Scordilis, 2006)

    Fig. 13 The results of regression analysis between Mw and mb.The solid lines show the results of regression and 1 standarddeviation

    J Seismol (2013) 17:645666 657

  • amplification at low frequencies have a strongercorrelation with the 30, compared with the am-plification at higher frequencies. A similar trendhas been seen for East-Central Iran by Hassani et

    al. (2011). It is clear also that site response in theAlborz region, Northern Iran, has a weak correla-tion with the 30, compared with the two otherregions.

    100 101101

    100

    101

    AM

    PLIF

    ICA

    TIO

    N 1Asr

    100 101101

    100

    1012Arn

    100 101101

    100

    1013Ann

    100 101101

    100

    1014Bah

    100 101101

    100

    101

    AM

    PLIF

    ICA

    TIO

    N 5Be1

    100 101101

    100

    1016Be2

    100 101101

    100

    1017Ber

    100 101101

    100

    1018Btn

    100 101101

    100

    101

    AM

    PLIF

    ICA

    TIO

    N 9Bod

    100 101101

    100

    10110Cai

    100 101101

    100

    10111CCn

    100 101101

    100

    10112Dsh

    100 101101

    100

    101

    AM

    PLIF

    ICA

    TIO

    N 13Dhr

    100 101101

    100

    10114DAa

    100 101101

    100

    10115Dbn

    100 101101

    100

    10116Drd

    100 101101

    100

    101

    AM

    PLIF

    ICA

    TIO

    N 17Gmh

    100 101101

    100

    10118Gom

    100 101101

    100

    10119HAd

    100 101101

    100

    10120Jhm

    100 101101

    100

    101

    Frequency(Hz)

    AM

    PLIF

    ICA

    TIO

    N 21Jum

    100 101101

    100

    101

    Frequency(Hz)

    22Krn

    100 101101

    100

    101

    Frequency(Hz)

    23Kes

    100 101101

    100

    101

    Frequency(Hz)

    24Ken

    Fig. 14 Site-amplification spectra at 40 stations obtained from performed inversion (solid), arrived from H/V method (dashed), and 1standard deviation range (shaded)

    658 J Seismol (2013) 17:645666

  • Already, Dutta et al. (2003) have reported a similartrend for S wave site response in Anchorage, Alaska,from weak-motion data using the GI method. Theresults of Fig. 17 show that the transition from soilto rock category (dashed line along the ordinate) cor-responds to amplification factor of 2.2 and 3, re-spectively at low and high frequencies.

    4.3 Residuals

    If a prediction model is not biased at any fixedpoint (i.e., at a specific frequency, distance, ormagnitude), the residuals between the observedand predicted values should be distributed normal-ly with a mathematical expectation of zero andwith different variances. Here, we examined thecomputed residuals, defined as the natural log of

    observed FAS minus the natural log of predictedFAS, as a function of distance and magnitude, tobe sure that the results are not biased. Table 2summarizes the model parameters which are eitherchosen or obtained in the inversion. Residuals forfrequencies from 0.4 to 15 Hz over all records arecalculated using the model parameters as listed inTable 2. The distribution of residuals was plottedagainst distance and magnitude to explore possibletrends in the results. Figure 18a shows the histogramof the residuals which show a normal distribution with asmall variance of about 0.26. Figure 18b, c shows thesame residuals but as a function of hypocentral distanceand magnitude, respectively. As it can be concludedfrom Figs. 18 and 19, there is no apparent bias and trendin the residuals; all the residual plots examined show noobvious dependence of the scatter on distance or

    100 101101

    100

    101A

    MPL

    IFIC

    ATI

    ON 25Ken

    100 101101

    100

    10126Km1

    100 101101

    100

    10127Ljb

    100 101101

    100

    10128Ner

    100 101101

    100

    101

    AM

    PLIF

    ICA

    TIO

    N 29NAd

    100 101101

    100

    10130Qsm

    100 101101

    100

    10131Rci

    100 101101

    100

    10132Rgn

    100 101101

    100

    101

    AM

    PLIF

    ICA

    TIO

    N 33Shh

    100 101101

    100

    10134Sld

    100 101101

    100

    10135Snr

    100 101101

    100

    10136Sua

    100 101101

    100

    101

    Frequency(Hz)

    AM

    PLIF

    ICA

    TIO

    N 37Tal

    100 101101

    100

    101

    Frequency(Hz)

    38Tmn

    100 101101

    100

    101

    Frequency(Hz)

    39TAd

    100 101101

    100

    101

    Frequency(Hz)

    40Zdr

    Fig. 14 (continued)

    J Seismol (2013) 17:645666 659

  • magnitude and in conclusion, the model is fitting thedata rather well.

    5 Summary and conclusions

    S wave spectra of earthquakes recorded by theBHRC strong-motion stations at regional scale aremodeled as the product of source, propagation(including geometric and anelastic attenuation),and site effects. SVD algorithm is employed tosolve the inverse problem and retrieve these dif-ferent terms, because it provides a numericallyrobust solution to the least squares problem. Inthis study, 148 three-component records were usedwhich contain 35 earthquakes of magnitude M4.2to M6.2. These earthquakes were recorded at 40stations in the hypocentral distance range from10 to 100 km. We constrained the inversion usingthe H/V ratio at seven reference sites which have arelatively flat response.

    In the moment (M0) range of 1015NmM0

    1018Nm, accepting the similarity assumption, thepresent study of source parameters yielded therelation M0c3=3.311016Nm1 s3, where fc iscorner frequency. The Brune stress drop ()

    101 100 101101

    100

    101G

    ener

    aliz

    ed in

    vers

    ion

    1.13 Hz

    101 100 101101

    100

    1012.59 Hz

    101 100 101101

    100

    1014.05 Hz

    101 100 101101

    100

    1015.51 Hz

    101 100 101101

    100

    101

    Gen

    eral

    ized

    inve

    rsio

    n

    6.97 Hz

    101 100 101101

    100

    1018.43 Hz

    101 100 101101

    100

    101

    H/V method

    9.89 Hz

    101 100 101101

    100

    101

    H/V method

    11.35 Hz

    101 100 101101

    100

    101

    H/V method

    Gen

    eral

    ized

    inve

    rsio

    n

    12.81 Hz

    101 100 101101

    100

    101

    H/V method

    14.27 Hz

    Fig. 15 Comparisons of amplification spectra between the GI and direct spectral ratio that were averaged over ten frequency bands

    101 100 101101

    100

    101

    Ave

    rage

    site

    resp

    onse

    by

    GI m

    etho

    d

    Average site response by H/V method

    Zagros

    CentralEeastNorthern Iran

    1:1

    1:2

    2:1

    Fig. 16 The comparison of average site response calculated byinversion versus that obtained from the H/Vmethod. The dashedlines in the figure indicate 1:2 and 2:1 correspondence. Whitecircles the Zagros stations used in the current study, crossesNorthern Iranian stations obtained from Zafarani et al. (2012),and triangles the site effects estimated for the stations in theCentral-Eastern Iran by Hassani et al. (2011)

    660 J Seismol (2013) 17:645666

  • estimates for individual earthquakes range fromabout 1.4 to 35.0 MPa and the average value isaround 6.6 MPa for the magnitude range of eventsconsidered in this study. The 1 standard deviationof the stress drop parameter corresponds to 16.1and 3.5 MPa, respectively. Using data from west-ern and eastern parts of the Zagros region, thestress drop has been estimated as 7.1 and 5.9 MPafor these regions, respectively. There is no significantdifference between source characteristics of two subre-gions. This was expected because both subregionsbelong to the Zagros seismotectonic zone, havingan Interplate regime. The relation between Es andMo can be expressed as log(Es)=(4.680.061)+logM0. To reduce the trade-off effects, thedistance-dependent kappa factors were indepen-dently determined from the slope of the amplitude

    of Fourier acceleration spectrum at higher frequen-cies of generally more than 5 Hz. The kappa forhorizontal and vertical components was found tobe 0.043 and 0.024, respectively. Also, no attemptwas made to estimate the path effects in the cur-rent study; instead, a path model was adoptedfrom previous studies. The site spectrum resultedin the current study was compared with the H/Vmethod. Due to the shortcomings of the H/V ratiotechnique, we do not expect the H/V site responseestimates to perfectly match the general inversionresults in amplitude, although a better match inshape was obtained and generally, comparison ofresults reveals that the H/V ratios successfullymatch the shapes of the transfer functions obtainedby the inversion technique. A comparison of siteamplification values obtained by the inversion

    102 103101

    100

    101

    Soil Rock

    s(m/s)

    Am

    plifi

    catio

    nLow frequency band

    102 103101

    100

    101

    RockSoil

    s(m/s)

    High frequency band

    ln(Amp)=(0.410.18)ln(s)+(4.211.23)

    ln(Amp)=(0.480.18)ln(s)+(4.051.20)

    Fig. 17 The site response values at low (left) and high (right)frequency bands are plotted with respect to the average S wavevelocity of the uppermost 30 m (circles); solid line current study,dashed line East-Central Iran (Hassani et al. 2011), dotted line

    the Alborz region, Northern Iran (Zafarani et al. 2012). Thedotted vertical line marks the transition from soil to rock sitecondition as defined by Zare et al. (1999)

    Table 2 Seismological parame-ters as obtained from its calibra-tion in this study or adoptedfrom previous studies

    Parameter Parameter value

    Crustal shear-wave velocity 3.5 (km/s)

    Crustal density (g/cm3) 2.8

    Q f Q f =153f 0.83Geometric spreading R1 (R40 km), (R=0.40)1/2, and (R>40 km)Site amplification Generalized inversion results (this study)

    Source spectrum Generalized inversion results (this study)

    Kappa (parameter of high-cut filter (s)) 0.043

    J Seismol (2013) 17:645666 661

  • method shows that both methods practically repre-sent a same shape for site response spectra, but,amplification values obtained from H/V ratios forsome of the sites are different from those obtained fromthe general inversion, and also there is a weak correla-tion between two methods in higher frequencies.Finally, for two frequency bands, the average of site

    response spectra were computed and related to theshear-wave velocity in logarithmic scale. It can be con-cluded that site amplification at low frequencies have astronger correlation with the 30, compared with theamplification at higher frequencies. A similar trend hasbeen seen for East-Central Iran by Hassani et al. (2011).It was also concluded that site response in the Alborz

    2 1 0 1 20

    100

    200

    300

    400

    500

    600

    700

    800

    residual

    nu

    mbe

    r of v

    alue

    s

    101 1022

    0

    2

    hypocentral distance (Km)

    resid

    ual

    3.5 4 4.5 5 5.5 6 6.52

    0

    2

    Mw

    resid

    ual

    Fig. 18 Histogram of the residuals of the logarithms of Fourier amplitude spectra (left). Residuals as a function of hypocentral distanceand magnitude (right top and bottom), respectively

    101 1022

    0

    2

    ln(O

    bs/Si

    m) f=1.0 Hz

    101 1022

    0

    2

    ln(O

    bs/Si

    m) f=7.0 Hz

    3 3.5 4 4.5 5 5.5 6 6.5 72

    0

    2

    3 3.5 4 4.5 5 5.5 6 6.5 72

    0

    2

    3 3.5 4 4.5 5 5.5 6 6.5 72

    0

    2

    Mw101 102

    2

    0

    2

    ln(O

    bs/Si

    m)

    rhypo(Km)

    f=12.4 Hz

    f=1.0 Hz

    f=7.0 Hz

    f=12.4.0 Hz

    Fig. 19 Residuals of FAS at frequencies of 1.0, 7.0, and 12.4 Hz, plotted as a function of log10(hypocentral distance; left) andmagnitude (right)

    662 J Seismol (2013) 17:645666

  • region, Northern Iran, has a weak correlation with the30, compared with the East-Central Iran and Zagrosregions.

    Acknowledgment The authors acknowledge the Building andHousing Research Centre of Iran for providing them with theaccelerograms and shear-wave velocities used in the currentstudy.

    Appendix A

    Table 3 List of earthquakes used in this study

    E. No. yyyy/mm/dd hh:min Latitude(deg)

    Longitude(deg)

    Depth(km)

    Numberof records

    Magnitude Stations(distance in km)

    References

    mb Ms Mw

    1 1998/08/21 05:13 34.22 48.18 25 3 4.9 4.5 5 NAd(36) Bod(69) Asr(47) E2006

    2 1999/05/06 23:00 29.54 51.93 7 5 5.7 6.3 6.2 Gmh(48) Ken(28) Bah(29)Gom(56) Ken(26)

    E2006

    3 1999/05/06 23:13 29.43 51.93 10 3 5.2 5.7 5.7 Bah(18) Gom(64) Ken(36) E2006

    4 1999/05/21 19:17 29.37 51.99 15 4 4.2 3.3 4.2 Kes(25) Bah(18) Rgn(22)Rci(28)

    E2006

    5 1999/05/30 00:15 29.46 51.95 25 4 4.5 4.2 4.5 Rgn(34) Rci(34) Kes(33)Ken(39)

    E2006

    6 1999/05/31 19:11 29.32 52.05 15 4 4.3 3.8 4.3 Rgn(19) Rci(28) Kes(24)Bah(19)

    E2006

    7 1999/06/11 03:05 29.37 51.99 25 4 4.4 3.9 4.4 Rgn(30) Rci(34) Ken(45)Kes(32)

    E2006

    8 1999/10/31 15:09 29.37 51.85 15 4 4.9 4.9 5.2 Gmh(60) Rgn(34) Ken(36)Bah(19)

    E2006

    9 1999/12/05 00:06 29.52 51.77 15 3 4.6 4.4 4.6 Ner(21) Bah(34) Rgn(44) E2006

    10 2002/04/24 19:43 34.53 47.36 25 5 4.8 4.6 4.8 Ann(26) Shh(39) Snr(43)Btn(31) Ljb(46)

    E2006

    11 2002/04/24 19:48 34.60 47.40 25 6 5.3 5.1 5.4 Ljb(40) Btn(35) Snr(37)Shh(39) Ann(25) Arn(58)

    E2006

    12 2002/04/24 20:10 34.49 47.35 25 4 4.8 4.3 4.8 Ann(28) Arn(58) Shh(39)Snr(47)

    E2006

    13 2002/04/24 20:11 34.53 47.16 25 3 4.5 0 4.5 Snr(55) Shh(54) Ann(32) E2006

    14 2002/05/17 15:52 29.48 51.96 25 3 4.8 4.2 4.8 Ken(37) Bah(33) Ner(26) E2006

    15 2002/12/24 17:03 34.56 47.48 20 6 5.1 4.6 5.2 Ann(24) Arn(48) Btn(28)Shh(29) Snr(34) Ljb(44)

    E2006

    16 2002/12/24 22:17 34.49 47.40 28 3 4.6 0 4.6 Snr(46) Shh(38) Ann(31) E2006

    17 2003/07/10 17:06 28.31 54.17 10 4 5.8 5.5 5.8 Zdr(60) Jum(21) Jhm(64)HAd(27)

    E2006

    18 2003/07/10 17:40 28.25 54.08 15 4 5.7 5.4 5.7 HAd(38) Jhm(61) Jum(18)Zdr(63)

    E2006

    19 2003/11/28 23:19 28.33 54.05 25 3 5.2 4.3 5 HAd(44) Jum(27) Dbn(30) E2006

    20 2005/11/27 10:22 26.77 55.90 12 6 6.1 5.8 5.9 Be2(62) Qsm(45) Krn(61)Ber(39) Be1(62) Sua(21)

    HRVD

    21 2006/03/30 16:17 33.55 48.71 10 3 4.8 4.8 CCn(25) Cai(22) Bod(39 NEIC

    22 2006/03/30 19:36 33.65 48.78 20 8 5.2 5.1 CCn(24) Bod(34) Drd(37)Km1(47) Cai(29) Sld(67)TAd(31) Dhr(40)

    HRVD

    23 2006/03/31 01:17 33.62 48.91 12 9 5.7 6 6.1 CCn(13) Cai(35) Km1(54)Drd(23) Sld(56) Dhr(26)TAd(38) NAd(101) Asr(67)

    HRVD

    24 2006/03/31 01:31 33.71 48.65 10 4 4.7 4.7 NEIC

    J Seismol (2013) 17:645666 663

  • Appendix B

    Table 3 (continued)

    E. No. yyyy/mm/dd hh:min Latitude(deg)

    Longitude(deg)

    Depth(km)

    Numberof records

    Magnitude Stations(distance in km)

    References

    mb Ms Mw

    CCn(27) Bod(24) Km1(38)Cai(14)

    25 2006/03/31 11:54 33.73 48.75 26 6 4.9 4.3 5.1 Km1(52) Sld(78) TAd(31)Drd(47) Bod(32) Cai(33)

    HRVD

    26 2006/04/02 01:59 33.76 48.91 15 3 4.2 4.2 DAa(22) Dsh(18) CCn(19) NEIC

    27 2006/04/12 11:47 33.72 48.79 11 3 4.6 4.6 DAa(21) CCn(17) Dsh(11) NEIC

    28 2006/06/28 21:02 26.77 55.81 12 4 5.8 5.8 5.8 Tmn(13) Ber(33) Qsm(52)Be2(68)

    HRVD

    29 2008/05/05 21:57 28.19 53.99 12 3 5.4 5.2 Dbn(33) Jum(14) Zdr(65) HRVD

    30 2008/09/10 11:00 26.65 55.72 12 5 6.1 6.1 6.1 Ber(38) Tmn(23) Sua(40)Tal(17) Krn(66)

    HRVD

    31 2008/09/11 02:16 26.93 55.63 7 3 5.2 5.2 Tmn(30) Sua(47) Tal(22) HRVD

    32 2008/09/17 17:43 26.75 55.96 12 3 5.3 5 5.2 Be1(60) Sua(17) Qsm(41) HRVD

    33 2008/12/07 13:36 26.82 55.74 12 4 5.7 5.4 5.4 Ber(25) Be1(70) Sua(35)Tal(14)

    HRVD

    34 2008/12/08 14:41 26.83 55.76 12 6 5.5 5.3 5.1 Be1(68) Sua(33) Tal(15)Qsm(54) Be1(68) Tal(15)

    HRVD

    35 2008/12/09 15:09 26.75 55.80 14 3 5.2 5 5 Sua(30) Qsm(55) Tal(16) HRVD

    Mw moment magnitude, E2006 Engdahl et al. (2006), HRVD Harvard Seismology, NEIC National Earthquake Information Centre

    Table 4 Tabulations of stations with their corresponding shear-wave velocity

    St. No. Station name Latitude (deg) Longitude (deg) Number of records V30 (m/s) Abbreviation

    1 Aleshtar 33.871 48.259 2 621 Asr

    2 Aran 34.41 47.92 3 Arn

    3 Armanijan 34.61 47.35 6 390 Ann

    4 Balaadeh 29.291 51.935 7 1,380 Bah

    5 Bandar-e-Abbas 1 27.193 56.293 5 337 Be1

    6 Bandar-e-Abbas 2 27.19 56.298 2 Be2

    7 Bandar-e-Khamir 26.952 55.582 4 679 Ber

    8 Bistoon 34.38 47.43 3 Btn

    9 Boroojerd 33.891 48.754 5 579 Bod

    10 Chaghalvandi 33.664 48.553 5 616 Cai

    11 Chalan Choolan 33.659 48.913 6 428 CCn

    12 Darbastaneh 33.701 48.817 2 1,103 Dsh

    13 Darreh-Asbar 33.45 49.06 2 935 Dhr

    14 Deh Azna 33.611 48.929 2 450 DAa

    15 Doobaran 28.411 54.182 2 1,363 Dbn

    16 Dorood 33.491 49.059 3 771 Drd

    17 Ghaemiyeh 29.846 51.59 2 617 Gmh

    18 Gooyom 29.829 52.4 2 598 Gom

    19 Haji Abad 28.35 54.42 3 561 HAd

    20 Jahrom 28.503 53.554 2 Jhm

    664 J Seismol (2013) 17:645666

  • References

    Aki K (1967) Scaling law of seismic spectrum. J Geophys Res72:12171231

    Ambraseys NN, Melville CP (1982) A history of Persian earth-quakes. Cambridge University Press, London, p 219 pp

    Anderson J, Hough S (1984) A model for the shape of theFourier amplitude spectrum of acceleration at high fre-quencies. Bull Seism Soc Am 74:19691993

    Andrews DJ (1986) Objective determination of source parame-ters and similarity of earthquakes of different size, earth-quake source mechanics. American Geophysical Union,Washington, pp 25967

    Ansari A, Noorzad A, Zafarani H, Vahidifard H (2010) Correctionof highly noisy strongmotion records usingmodified waveletde-noising method. Soil Dyn Earthq Eng 30:1168181

    Berberian M (1995) Master blind thrust faults hidden underthe Zagros folds: active basement tectonics and surfacemorphotectonics. Tectonophysics 241:193224

    Boatwright J, Fletcher J, Fumal T (1991) A general inversionscheme for source, site, and propagation characteristicsusing multiply recorded sets of moderate-sized earth-quakes. Bull Seism Soc Am 81:17541782

    Boore DM (2003) Simulation of ground motion using the sto-chastic method. Pure Appl Geophys 160:635675

    Brune J (1970) Tectonic stress and the spectra of seismic shearwaves from earthquakes. J Geophys Res 75:49975009

    Brune J (1971) Correction to Brune (1970). J Geophys Res 76:5002Buiding and Housing Research Center (BHRC). Available from

    http://www.bhrc.ac.ir/portal/. Accessed May 2011Castro R, Anderson JG, Singh SK (1990) Site response, attenua-

    tion, and source spectra of S waves along the Guerrero,Mexico subduction zone. Bull Seism Soc Am 80:14811503

    Chen SZ, Atkinson GM (2002) Global comparisons of earth-quake source spectra. Bull Seism Soc Am 92:88595

    Drouet S, Chevrot S, Cotton F, Souriau A (2008) Simulta-neous inversion of source spectra, attenuation parame-ters, and site responses: application to the data of theFrench accelerometric network. Bull Seism Soc Am98:198219

    Dutta U, Martirosyan A, Biswas N, Papageorgiou A, CombellickR (2001) Estimation of S-wave site response in Anchorage,Alaska, from weak-motion data using generalized inversionmethod. Bull Seism Soc Am 91:335346

    Dutta U, Biswas N, Martirosyan A, Papageorgiou A, KinoshitaS (2003) Estimation of earthquake source parameters andsite response in Anchorage, Alaska from strong-motionnetwork data using generalized inversion method. PhysEarth Planet Inter 137:1329

    Engdahl ER, Jackson JA, Myers SC, Bergman EA, Priestley K(2006) Relocation and assessment of seismicity in the Iranregion. Geophys J Int 167:76178

    Harmsen SC (1997) Determination of site amplification in theLos Angeles Urban Area from inversion of strong-motionrecords. Bull Seism Soc Am 87:86687

    Table 4 (continued)

    St. No. Station name Latitude (deg) Longitude (deg) Number of records V30 (m/s) Abbreviation

    21 Jouyom 28.259 53.982 4 1,244 Jum

    22 Kahoorestan 27.216 55.564 2 807 Krn

    23 KarehBas 29.45 52.17 4 Kes

    24 Kazeroon 29.62 51.67 2 Ken

    25 Khan zeynioun 29.671 52.147 5 773 Ken

    26 Khoram Abad1 33.491 48.359 4 Km1

    27 Lenj Ab 34.87 47.28 3 Ljb

    28 Narges Zar 29.447 51.891 2 983 Ner

    29 Noor Abad 34.072 47.972 2 758 NAd

    30 Qeshm 26.962 56.275 5 757 Qsm

    31 Richi 29.5 52.18 4 1,050 Rci

    32 Romghan 29.371 52.162 6 1,362 Rgn

    33 Sahneh 34.47 47.68 6 Shh

    34 Shool Abad 33.184 49.192 3 1,084 Sld

    35 Sonqor 34.78 47.6 6 1,477 Snr

    36 Suza 26.782 56.07 7 1,334 Sua

    37 Tabl 26.758 55.725 6 931 Tal

    38 Tomban 26.766 55.863 3 778 Tmn

    39 Tooshk-e-Ab-e-Sard 33.773 48.569 3 891 TAd

    40 Zahedshahr 28.742 53.805 3 390 Zdr

    Stations set in italics are the selected reference sites

    J Seismol (2013) 17:645666 665

  • Hartzell S (1992) Site response estimation from earthquake data.Bull Seism Soc Am 82:230827

    Harvard Seismology (2011) Centroid Moment Tensor (CMT) cata-log search. www.seismology.harvard.edu/. AccessedMay 2011

    Hassani B, Zafarani H, Farjoodi J, Ansari A (2011) Esti-mation of site amplification, attenuation and sourcespectra of S-waves in the East-Central Iran. Soil DynEarthq Eng 31:13971413

    Husid P (1967). Gravity effects on the earthquake response ofyielding structures. Report of Earthquake Engineering Re-search Laboratory, California Institute of Technology,Pasadena, California

    Iwata T, Irikura K (1988) Source parameters of the 1983 JapanSea earthquake sequence. J Phys Earth 36:15584

    Lermo J, Chavez-Garcia F (1993) Site effect evaluation using spectralratios with only one station. Bull Seism Soc Am 83:157494

    Madariaga R (1976) Dynamics of an expanding circular fault.Bull Seism Soc Am 66:639666

    McCann MWJ, Shah HC (1979) Determining strong motion du-ration of earthquakes. Bull Seism Soc Am 69:12531265

    Mirzaei Alavijeh, H. & Farzanegan, E., 1998. Specifications ofthe Iranian Accelerograph Network Stations Building andHousing, Research Center, Publication No. B-280.

    Mirzaei Alavijeh H, Sinaiean F, Farzanegan E, Sadeghi AlavijehME (2007). Iran Strong Motion Network (ISMN) prospectsand achievements. In: Proceedings of the fifth internationalconference on seismology and earthquake engineering, Tehran

    Motazedian D (2006) Region-specific key seismic parameters forearthquakes in Northern Iran. Bull Seism Soc Am 96:138395

    Mousavi M, Zafarani H, Noorzad A, Ansari A, Bargi KH (2007)Analysis of Iranian strong motion data using the specificbarrier model. J Geophys Eng 4:114

    National Earthquake Information Centre (NEIC). Available fromhttp://earthquake.usgs.gov/regional/neic/. Accessed May 2011

    Nowroozi A (2010) Probability of peak ground horizontal andpeak ground vertical accelerations at Tehran and surround-ing areas. Pure Appl Geophys 167:14591474

    Ramsey LA, Walker RT, Jackson J (2008) Fold evolution anddrainage development in the Zagros Mountains of Farsprovince, SE Iran. Basin Research 20:2348

    Salazar W, Sardina V, Cortina JD (2007) A hybrid inversiontechnique for the evaluation of source, path, and site effectsemploying S-wave spectra for subduction and upper-crustalearthquakes in El Salvador. Bull Seism Soc Am 97:208221

    Scholz CH, Aviles CA, Wesnousky SG (1986) Scaling differ-ences between large interplate and intraplate earthquakes.Bull Seism Soc Am 76:6570

    Scordilis EM (2006) Empirical global relations converting MSand mb to moment magnitude. J Seismology 10:225236

    Shoja-Taheri J, Niazi M (1981) Seismicity of Iranian plateauand bordering regions. Bull Seism Soc Am 71:477489

    Shoji Y, Kamiyama M (2002) Estimation of local site effects bya generalized inversion scheme using observed records ofSmall-Titan. Soil Dyn Earthq Eng 22:855864

    Siddiqqi J, Atkinson GM (2002) Ground-motion amplificationat rock sites across Canada as determined from thehorizontal-to-vertical component ratio. Bull Seism SocAm 92:877884

    Silva WJ, Abrahamson N, Toro G, Costantino C (1997). De-scription and validation of the stochastic ground motionmodel. Final report. Brookhaven National Laboratory, As-sociated Universities, Inc., Upton, New York

    Sokolov VY, Bonjer KP, Oncescu M, Rizescu M (2005) Hardrock spectral models for intermediate-depth Vrancea(Romania) earthquakes. Bull Seism Soc Am 95:17491765

    Soghrat MR, Khaji N, Zafarani H (2012) Simulation of strongground motion in northern Iran using the specific barriermodel. Geophys J Int 188:645679

    Talebian M, Jackson J (2004) A reappraisal of earthquake focalmechanisms and active shortening in the Zagros mountainsof Iran. Geophys J Int 156:121

    Trifunac MD, Lee VW (1973). EERL 73-03, Earthquake Engi-neering Research Laboratory, California Institute of Tech-nology Pasadena, 1973-10 (370/T72/1973). 360 pp

    Tramelli A, Galluzzo D, Del Pezzo E, Di VitoMA (2010) A detailedstudy of the site effects in the volcanic area of Campi Flegreiusing empirical approaches. Geophys J Int 182:10731086

    Vassiliou MS, Kanamori H (1982) The energy release in earth-quake. Bull Seism Soc Am 72:371387

    Vernant F, Nilforoushan F, Hatzfeld D, Abbassi MR, Vigny C,Masson F, Bayer R, Nankali H, Martinod J, Ashtiani A,Tavakoli RF, Chery J (2004) Present-day crustal deforma-tion and plate kinematics in the Middle East constrained byGPS measurements in Iran and northern Oman. Geophys JInt 157:381398

    Zafarani H, Mousavi M, Noorzad A, Ansari A (2008) Calibra-tion of the specific barrier model to Iranian plateau earth-quakes and development of physically based attenuationrelationships for Iran. Soil Dyn Earthq Eng 28:55076

    Zafarani H, Noorzad A, Ansari A, Bargi K (2009) Stochasticmodeling of Iranian earthquakes and estimation of groundmotion for future earthquakes in Greater Tehran. Soil DynEarthq Eng 29:72241

    Zafarani H, Hassani B, Ansari A (2012) Estimation of earthquakeparameters in the Alborz seismic zone, Iran using generalizedinversion method. Soil Dyn Earthq Eng 42:197218

    Zafarani H, Soghrat MR (2012) Simulation of ground motion inthe Zagros, Iran using the specific barrier model and sto-chastic method. Bull Seism Soc Am 102:20312045

    Zare M, Bard PY, Ghafory-Ashtiany M (1999) Site character-izations for the Iranian strong motion network. Soil DynEarthq Eng 18:10123

    666 J Seismol (2013) 17:645666

    Site response and source spectra of S waves in the Zagros region, IranAbstractIntroductionDataMethod of analysisInversion resultsSource modelSite responseResiduals

    Summary and conclusionsAppendix AAppendix BReferences