##
Kuzey Batı Anadolu cisim dalgalarının spektral özellikleri

Kuzey Batı Anadolu cisim dalgalarının spektral özellikleri

##### Dosyalar

##### Tarih

1990

##### Yazarlar

Ergin, Mehmet

##### Süreli Yayın başlığı

##### Süreli Yayın ISSN

##### Cilt Başlığı

##### Yayınevi

Fen Bilimleri Enstitüsü

##### Özet

Bu tezde Marmara bölgesinde 39 - 41 °N enlemleri İle 26 - 31 °E boylamları arasında 1987 Eylül - 1989 Mayıs döneminde oluşmuş, TÜBİTAK Gebze sayısal kayıt istasyonu tarafından kayıt edilmiş olan 37 tane deprem sismogramı değerlendirilerek, kaynak parametreleri saptanmış, elastik cisim dalgalarının spektral özellikleri incelenerek bölgesel farklılıkların saptanmasına çalışılmıştır. Belli bölgelerdeki yöresel özellikleri incelemek amacıyla episantırları birbirlerine yakın olan depremler kümeler halinde incelenmiştir. Bu depremlerin manyitüdleri sinyal devam sürelerinden yararlanılarak hesaplanmıştır. İncelenen depremlerin manyitüdlerinin 2.2 ile 4.2 arasında değişmektedir. Seçilen depremlerin düşey, radyal ve transvers bileşen hız sismogramlarından yer değiştirme sismogramlarma geçildikten sonra spektrumlardan alet etkisi giderilip, Brune (1970,1971) 'un dairesel kaynak modeli benimsenerek yer hareketi spektrumlarından kaynak parametreleri hesaplanmıştır. Tek istasyon verileri kullanıldığından yayınım pattern ve soğurma etkileri dikkate alınmamıştır. P dalgası düşey ve radyal bileşen için sismik moment ortalama 1.55 10XB dyn.cm, dalga yayılım enerjisi 4.5 İO11 erg ve gerilme düşümü ise 1 bar 'dan küçük olarak bulunmuştur. Aynı şekilde, S dalgası için bulunan sismik moment ortalama 1.158 103-9 dyn.cm, dalga yayılım enerjisi 2.378 10X2 - erg ve gerilme düşümü ise 1 bar 'dan küçük degerindedirler. P dalgaları için bulunmuş olan kaynak yarıçapları ortalama 692 metre civarında olup S dalgalarınkine göre yaklaşık iki katı büyüklüktedir. Spektral analiz sonucunda elde edilen kaynak parametrelerinin birbirleriyle olan ilşkileri çıkarılmaya çalışılmış ancak veri sayısının yetersiz oluşu nedeniyle, bunların istatistiksel analizini yapmak mümkün olmamıştır. Gözlenmiş spektrumlardaki benzerliklerden yola çıkarak bölgesel farklılıkları ortaya çıkarmak amacıyla spektrumlar normalize edilerek yığılmışlardır, özellikle Gebze istasyonunun güneyinde kalan 39 - 39.5 °N enlem ve 27.79 - 30.15 °E boylamlarında yer alan depremlerin yığılmış spektrumlarında düşük frekanslarda genelde genliklerin aşağıya çekilmiş oldukları ve buna da muhtemel band sınırlı bir soğrulma olayının neden olduğu düşünülmektedir.

For a better understanding of earthquake source processes, seismologists conduct various studies through modeling. Such studies can be classified as kinematic and dynamic; kinematic is the branch of mechanics that deals purely with motion, without analyzing the underlying forces that cause or participate in the motion and in practice it enables us to interpret the observable motions that radiate from the source region in terms of particle motions on a fault plane, dynamics is the branch of mechanics that deals directly with force systems, and with the energy balance that governs motion. To understand the physical processes actually occurring in the source region, one must study stress-depended material properties. That is, one examines the way in which material failure nucleates and spreads, rapidly relieving stresses that h«d slowly risen to exceed the strength of material in the source region (Aki, 1980). Progress in the study of focal mechanisms in the last ten years has led to the definition and determination of several parameters describing more fully processes at the source of earthquakes, such as the fracture dimensions, dislocation, seismic moment, average stress and stress drop. Dislocation dynamic models of the source of earthquakes predict the shapes of the spectrum of seismic waves for the far field displacement. From the parameters which define the form of such spectrum certain important characteristics of the dimensions and stress conditions at the source can be determined. Many authors have in the recent years treated this problem, among them, Aki (1967,1972) Ben Menahem (1961,1965), Brune (1970), Hanks and Thatcher (Î972), Hanke and Wyss (1970) may be mentioned. Recent studies have successfully applied dislocation' theory to the study of near-source displacements. Unfortunately the dislocation models used have been somewhat arbitrary in the specification of the time function of the dislocation motion. Aki arbitrarily assumed the dislocation occurred as a step function in time, whereas Haskell assumed a rump function with slope estimated from the expected duration fault slippage. In this study, the time function is related to the effective stress available to accelerate the two sides of the fault as the difference between the initial stress «i and a V dynamic ' frictional stress1 «v, which is of the opposite sense and always acts to resist the fault slip; i.e., 0=0,-0,. The results provide a physical basis for integration methods such as those of Aki and Haskell and in addition provide a basis for understanding the time function and spectrum at high frequencies, which are of particular interest in engineering seismology and in the study of earthquake spectra. The observations made in this study are interpreted in terms of Brune's theory, and so his source model is briefly considered here. This includes the following assumptions; the rupture area is assumed to be flat a circle, the stress is applied instantaneously, uniformly (no propagation effect) and tangentially to the interior of the dislocation surface, during the rupture the fault surface is totally reflective for elastic waves, the stress drop during rupture is complete and finally the particle situated at the center of the fault has an exponentially decreasing velocity due to the f initeness of the fault. In the model of Brune's (1970) tangential stress step is applied to the interior of a dislocation surface causing the fault block on one side to move in one direction, and the block on the other side, in the opposite direction. The step is assumed to apply instantaneously over the fault surface,- that is to say the fault propagation effects are neglected, for simplicity. Also the elastic events on the two sides of the fault are isolated from one another by fault surface. The stress step sends a pure shear stress wave propagating perpendicularly to the fault surface. At a point near the fault, therefore, displacement increases linearly in time as the stress pulse propagates away from the fault. Then it levels off as the finiteness of the faults becomes felt at the observation point from the fault's extremity, and is longest at the center of the fault. The particle velocity will begin to decrease to zero when the effects of the fault finiteness reach the observation point from the fault's extremities. Brune introduced a time constant, *, equivalent to the travel time of this signal, r=o/^ t where a is the equivalent radius of the fault surface. The effects of the finite velocity of rupture propagation are suitably averaged over the azimuth. The Haskell and Brune models are not equally applicable to all practical problems. The Haskell model describes the effects of rupture propagation well, but the use of a ramp-type source function is rather arbitrary and makes further analysis difficult in comparison with the case of an analytical function..The Brune model, on the other hand, employs an exponential function, so that mechanical consideration in the vicinity of a fault my be dealt with relatively easily. However, it does not consider explicitly rupture propagation with a finite velocity, in other words, it does not apply well to far-field problems vi in which the effect of rupture propagation is of special concern. To summarize these models are complementary in application, and the most appropriate model must be chosen depending on whether far-field or near-field problems, respectively, are to be studied. Most often amplitude spectra of seismic waves consist of two features: a flat low- frequency region and followed by a high frequency asymptote. The transition occurs at a frequency to which we will refer hereafter as the corner frequency. Those spectra can be described by three parameters: the average low frequency amplitude level ( n* ), the corner frequency ( f o ) and the power of the high frequency asymptote « (Brune, 1970). The seismic source parameters, seismic moment Mo, source dimension r, shear-stress drop *«, effective Bhear stress «?«, radiated energy Es, and apparent stress «?* can all be expressed in terms of this three spectral parameters that specify the far-field displacement of J. N. Brune' s 1970 seismic source model. All the above source parameters can be easily extracted from a log-log plot of a. versus frequency, For a particular case when « = 1, the spectral amplitudes decay approximately as the inverse of the square of frequency. This case represents a dislocation in which the stress drop -is complete (Hanks and Thatcher, 1972). These three parameters of the spectrum can be related to source characteristics. The spectral amplitude level at low frequencies &<. is related to the seismic moment Mo, and the corner frequency f0 is related to the dimensions of the fault, and « with the ratio of the stress drop to the effective stress. In principle spectra from either body or surface waves can be used in the determination of source parameters. Seismic moments are generally obtained from spectra derived; from well-dispersed surface waves. The preference for surface wave spectra is that, for large-shallow earthquakes, spectral information at periods in the range of several hundred seconds can readily be obtained; for body phases, the long period spectral data will be contaminated by multiple arrivals that follow within 60 to 100 sec, except at very restricted ranges of depths,and epicentral distances, on the other hand, the smaller and deeper earthquakes, generate significantly smaller surface waves, and the use öf, body wave spectra in the moment determinations for these events is preferable. Again with the exception of the larger shallow earthquakes, body wave spectra are also preferable for the determination of the source dimension, since fD is generally in a period range at which surface waves amplitudes area sensitive function of the propagation path (Hanks and Wyss, 1972). The purpose of this work is to develop a simple scheme for estimating seismic source parameters in terms of the three independent spectral parameters that specify the far-field displacement spectra that result from the model vii of the seismic source developed by Brune (1970) and also to investigate some regional variations of estimated parameters of 37 earthquakes. This estimation of relationship based on between distance and corner frequency, fault area and magnitude, seismic moment and fault dimension, stress drop and seismic moment and energy and magnitude. This comparison based on P-waves on vertical component, P-waves on radial component and S-waves on transverse component. The data is recorded at TÜBİTAK Gebze Digital Seismological Station on three component short period (fN = 1 Hz) digital seismometer which is 25 sample/s, and has a dynamic range 78 dB. Epicenters are taken from National Earthquake Information Center Catalog (NEICC). These earthquakes occurred 39 - 41°N latitudes and 26 - 31°E longitudes (Southern Marmara Region) between September 1987 - May 1989. Events are classified into 8 groups according to their epicentral distribution. The epicentral distances rang between 30 and 270 km. Magnitudes are assessed from signal duration and lie between 2.2 and 4.9. Having chosen the data the two horizontal components are rotated into the radial and tangential directions. We than separate the phase that we are interested in. Then we go to the signal conditioning. First the displacement seismograms have been obtained from velocity seismograms by trapezoidal integration rule. All spectra have been corrected for instrument response by using the response curve of the instrument. DC level has been removed, and a cosine taper was applied to both ends of the time window. The Fast Fourier Transform was used to obtain the spectra. No radiation pattern and attenuation correction were applied since single station data was used. Vertical component and radial of P waves along with the transverse components of shear waves were spectrally analyzed. The long period levels and corner frequencies of the spectra were determined from the low and high frequency asymptotes. The seismic source parameters of the earthquakes have been estimated according to J. N. Brune' s 1970 seismic source model assuming that *> = 2.7 g/cm3, P-wave velocity ° = 6.3 km. /sec, S-wave velocity £ = 3.5 km. /sec. Under Brune' s circular model assumptions, all spectra show a well defined corner frequencies (fo) and constant level of spectral amplitudes a*. The corner frequencies of the vertical component of P waves are in the range of 2.5 - 5.0 Hz. The seismic moment and the radius of the circular fracture in the source region. have been calculated following Keilis-Borok (1960) and Brune (1970). Seismic moments range from lO1-7 to 1020 dyn.cm and the average radius of all events is 700 m. By using seismic moments and fault radius the corresponding stress drops are found and less than 1 bar in general. It vli'i a apparently increases with increasing seismic moment. Radiated energies are also found within the range of 10 to 103-2 erg. Since vertical component of P-wave have very similar spectral properties, the spectral and source parameters obtained from them are similar to those found from radial components. It has been observed that the spectra of the transverse component of S-wave have higher corner frequency levels and steeper slopes of high frequency asymptotes compared to P waves spectra. The corner frequencies of the transverse component of S waves are in the range of 2.8 - 5.0 Hz and the seismic moments are found between 10xa and 1019 dyn.cm. The stress drops are in general less than 1 bar and apparently increase with increasing seismic moment. Calculated source radius is about 350 m. which is smaller than that of P-waves components. The radiated energy of the S-wave range from 103-0 to 103-3 erg. Empirical relations between magnitude and fault area, stress drop and seismic moment are valid; but the one relating energy and magnitude is not verified since data is very much scattered. Due to the sparsity of the data, it was impossible to arrive at meaningful statistical results. Characteristic parameters can be attributed to each region; however this is not attempted due to the insufficiency of data size. It was observed that the seismic source signals have been effected by attenuation at low frequencies. This phenomenon which is also clearly illustrated by stacking the data may be associated with a possible existence of a regional attenuation zone at lower crustal levels. This constitutes a preliminary work. Because; a number of events which used here is, not very large, the magnitude don't cover a very wide range, we don't have much idea about propagation media and all the data was collected from single station.

For a better understanding of earthquake source processes, seismologists conduct various studies through modeling. Such studies can be classified as kinematic and dynamic; kinematic is the branch of mechanics that deals purely with motion, without analyzing the underlying forces that cause or participate in the motion and in practice it enables us to interpret the observable motions that radiate from the source region in terms of particle motions on a fault plane, dynamics is the branch of mechanics that deals directly with force systems, and with the energy balance that governs motion. To understand the physical processes actually occurring in the source region, one must study stress-depended material properties. That is, one examines the way in which material failure nucleates and spreads, rapidly relieving stresses that h«d slowly risen to exceed the strength of material in the source region (Aki, 1980). Progress in the study of focal mechanisms in the last ten years has led to the definition and determination of several parameters describing more fully processes at the source of earthquakes, such as the fracture dimensions, dislocation, seismic moment, average stress and stress drop. Dislocation dynamic models of the source of earthquakes predict the shapes of the spectrum of seismic waves for the far field displacement. From the parameters which define the form of such spectrum certain important characteristics of the dimensions and stress conditions at the source can be determined. Many authors have in the recent years treated this problem, among them, Aki (1967,1972) Ben Menahem (1961,1965), Brune (1970), Hanks and Thatcher (Î972), Hanke and Wyss (1970) may be mentioned. Recent studies have successfully applied dislocation' theory to the study of near-source displacements. Unfortunately the dislocation models used have been somewhat arbitrary in the specification of the time function of the dislocation motion. Aki arbitrarily assumed the dislocation occurred as a step function in time, whereas Haskell assumed a rump function with slope estimated from the expected duration fault slippage. In this study, the time function is related to the effective stress available to accelerate the two sides of the fault as the difference between the initial stress «i and a V dynamic ' frictional stress1 «v, which is of the opposite sense and always acts to resist the fault slip; i.e., 0=0,-0,. The results provide a physical basis for integration methods such as those of Aki and Haskell and in addition provide a basis for understanding the time function and spectrum at high frequencies, which are of particular interest in engineering seismology and in the study of earthquake spectra. The observations made in this study are interpreted in terms of Brune's theory, and so his source model is briefly considered here. This includes the following assumptions; the rupture area is assumed to be flat a circle, the stress is applied instantaneously, uniformly (no propagation effect) and tangentially to the interior of the dislocation surface, during the rupture the fault surface is totally reflective for elastic waves, the stress drop during rupture is complete and finally the particle situated at the center of the fault has an exponentially decreasing velocity due to the f initeness of the fault. In the model of Brune's (1970) tangential stress step is applied to the interior of a dislocation surface causing the fault block on one side to move in one direction, and the block on the other side, in the opposite direction. The step is assumed to apply instantaneously over the fault surface,- that is to say the fault propagation effects are neglected, for simplicity. Also the elastic events on the two sides of the fault are isolated from one another by fault surface. The stress step sends a pure shear stress wave propagating perpendicularly to the fault surface. At a point near the fault, therefore, displacement increases linearly in time as the stress pulse propagates away from the fault. Then it levels off as the finiteness of the faults becomes felt at the observation point from the fault's extremity, and is longest at the center of the fault. The particle velocity will begin to decrease to zero when the effects of the fault finiteness reach the observation point from the fault's extremities. Brune introduced a time constant, *, equivalent to the travel time of this signal, r=o/^ t where a is the equivalent radius of the fault surface. The effects of the finite velocity of rupture propagation are suitably averaged over the azimuth. The Haskell and Brune models are not equally applicable to all practical problems. The Haskell model describes the effects of rupture propagation well, but the use of a ramp-type source function is rather arbitrary and makes further analysis difficult in comparison with the case of an analytical function..The Brune model, on the other hand, employs an exponential function, so that mechanical consideration in the vicinity of a fault my be dealt with relatively easily. However, it does not consider explicitly rupture propagation with a finite velocity, in other words, it does not apply well to far-field problems vi in which the effect of rupture propagation is of special concern. To summarize these models are complementary in application, and the most appropriate model must be chosen depending on whether far-field or near-field problems, respectively, are to be studied. Most often amplitude spectra of seismic waves consist of two features: a flat low- frequency region and followed by a high frequency asymptote. The transition occurs at a frequency to which we will refer hereafter as the corner frequency. Those spectra can be described by three parameters: the average low frequency amplitude level ( n* ), the corner frequency ( f o ) and the power of the high frequency asymptote « (Brune, 1970). The seismic source parameters, seismic moment Mo, source dimension r, shear-stress drop *«, effective Bhear stress «?«, radiated energy Es, and apparent stress «?* can all be expressed in terms of this three spectral parameters that specify the far-field displacement of J. N. Brune' s 1970 seismic source model. All the above source parameters can be easily extracted from a log-log plot of a. versus frequency, For a particular case when « = 1, the spectral amplitudes decay approximately as the inverse of the square of frequency. This case represents a dislocation in which the stress drop -is complete (Hanks and Thatcher, 1972). These three parameters of the spectrum can be related to source characteristics. The spectral amplitude level at low frequencies &<. is related to the seismic moment Mo, and the corner frequency f0 is related to the dimensions of the fault, and « with the ratio of the stress drop to the effective stress. In principle spectra from either body or surface waves can be used in the determination of source parameters. Seismic moments are generally obtained from spectra derived; from well-dispersed surface waves. The preference for surface wave spectra is that, for large-shallow earthquakes, spectral information at periods in the range of several hundred seconds can readily be obtained; for body phases, the long period spectral data will be contaminated by multiple arrivals that follow within 60 to 100 sec, except at very restricted ranges of depths,and epicentral distances, on the other hand, the smaller and deeper earthquakes, generate significantly smaller surface waves, and the use öf, body wave spectra in the moment determinations for these events is preferable. Again with the exception of the larger shallow earthquakes, body wave spectra are also preferable for the determination of the source dimension, since fD is generally in a period range at which surface waves amplitudes area sensitive function of the propagation path (Hanks and Wyss, 1972). The purpose of this work is to develop a simple scheme for estimating seismic source parameters in terms of the three independent spectral parameters that specify the far-field displacement spectra that result from the model vii of the seismic source developed by Brune (1970) and also to investigate some regional variations of estimated parameters of 37 earthquakes. This estimation of relationship based on between distance and corner frequency, fault area and magnitude, seismic moment and fault dimension, stress drop and seismic moment and energy and magnitude. This comparison based on P-waves on vertical component, P-waves on radial component and S-waves on transverse component. The data is recorded at TÜBİTAK Gebze Digital Seismological Station on three component short period (fN = 1 Hz) digital seismometer which is 25 sample/s, and has a dynamic range 78 dB. Epicenters are taken from National Earthquake Information Center Catalog (NEICC). These earthquakes occurred 39 - 41°N latitudes and 26 - 31°E longitudes (Southern Marmara Region) between September 1987 - May 1989. Events are classified into 8 groups according to their epicentral distribution. The epicentral distances rang between 30 and 270 km. Magnitudes are assessed from signal duration and lie between 2.2 and 4.9. Having chosen the data the two horizontal components are rotated into the radial and tangential directions. We than separate the phase that we are interested in. Then we go to the signal conditioning. First the displacement seismograms have been obtained from velocity seismograms by trapezoidal integration rule. All spectra have been corrected for instrument response by using the response curve of the instrument. DC level has been removed, and a cosine taper was applied to both ends of the time window. The Fast Fourier Transform was used to obtain the spectra. No radiation pattern and attenuation correction were applied since single station data was used. Vertical component and radial of P waves along with the transverse components of shear waves were spectrally analyzed. The long period levels and corner frequencies of the spectra were determined from the low and high frequency asymptotes. The seismic source parameters of the earthquakes have been estimated according to J. N. Brune' s 1970 seismic source model assuming that *> = 2.7 g/cm3, P-wave velocity ° = 6.3 km. /sec, S-wave velocity £ = 3.5 km. /sec. Under Brune' s circular model assumptions, all spectra show a well defined corner frequencies (fo) and constant level of spectral amplitudes a*. The corner frequencies of the vertical component of P waves are in the range of 2.5 - 5.0 Hz. The seismic moment and the radius of the circular fracture in the source region. have been calculated following Keilis-Borok (1960) and Brune (1970). Seismic moments range from lO1-7 to 1020 dyn.cm and the average radius of all events is 700 m. By using seismic moments and fault radius the corresponding stress drops are found and less than 1 bar in general. It vli'i a apparently increases with increasing seismic moment. Radiated energies are also found within the range of 10 to 103-2 erg. Since vertical component of P-wave have very similar spectral properties, the spectral and source parameters obtained from them are similar to those found from radial components. It has been observed that the spectra of the transverse component of S-wave have higher corner frequency levels and steeper slopes of high frequency asymptotes compared to P waves spectra. The corner frequencies of the transverse component of S waves are in the range of 2.8 - 5.0 Hz and the seismic moments are found between 10xa and 1019 dyn.cm. The stress drops are in general less than 1 bar and apparently increase with increasing seismic moment. Calculated source radius is about 350 m. which is smaller than that of P-waves components. The radiated energy of the S-wave range from 103-0 to 103-3 erg. Empirical relations between magnitude and fault area, stress drop and seismic moment are valid; but the one relating energy and magnitude is not verified since data is very much scattered. Due to the sparsity of the data, it was impossible to arrive at meaningful statistical results. Characteristic parameters can be attributed to each region; however this is not attempted due to the insufficiency of data size. It was observed that the seismic source signals have been effected by attenuation at low frequencies. This phenomenon which is also clearly illustrated by stacking the data may be associated with a possible existence of a regional attenuation zone at lower crustal levels. This constitutes a preliminary work. Because; a number of events which used here is, not very large, the magnitude don't cover a very wide range, we don't have much idea about propagation media and all the data was collected from single station.

##### Açıklama

Tez (Yüksek Lisans) -- İstanbul Teknik Üniversitesi, Fen Bilimleri Enstitüsü, 1990

##### Anahtar kelimeler

Jeofizik Mühendisliği,
Cisim dalgaları,
Kuzeybatı Anadolu,
Spektrum analizi,
Geophysics Engineering,
Body waves,
Northwest Anatolia,
Spectrum analysis