Open Geosci. 2017; 9:480–490

Research Article

Fadime Sertcelik* and Mehmet Guleroglu Coda Wave Attenuation Characteristics for North Anatolian Fault Zone, https://doi.org/10.1515/geo-2017-0037 Keywords: Coda wave attenuation, Frequency dependent, Received Dec 07, 2016; accepted Aug 03, 2017 North Anatolian Fault Zone, Lapse time Abstract: North Anatolian Fault Zone, on which large have occurred in the past, migrates regu- larly from east to west, and it is one of the most ac- 1 Introduction tive faults in the world. The purpose of this study is to estimate the coda wave quality factor (Qc) for each of Seismic attenuation express as the decrease in the energy the five sub regionsthat were determined according to that seismic waves propagate in the earth and it is similar the fault rupture of these large earthquakes and along to seismic velocities. The variance in attenuation charac- the fault. 978 records have been analyzed for 1.5, 3, 6, teristics on the crust is a function of depth and lateral tran- 9, 12 and 18 Hz frequencies by Single Backscattering sition. Generally, the change in attenuation in the earth is Method. Along the fault, the variations in the Qc with a bit greater than the change in velocity. The energy de- lapse time are determined via, Qc = (136±25)f(0.96±0.027), crease for a seismic wave is directly proportional to ampli- Qc = (208±22)f(0.85±0.02)Qc = (307±28)f(0.72±0.025) at 20, 30, tude, and inversely proportional to the square of the dis- 40 sec lapse times, respectively. The estimated average tance. If the decrease in energy is arisen from the lateral frequency-dependence quality factor for all lapse time reflection, fracture, and/or scattering, it is defines asthe are; Qc(f) = (189±26)f(0.86±0.02) for Karlıova-Tokat region; attenuation of scattering (Qs), and if it is due to its convert- Qc(f) = (216±19)f(0.76±0.018) for Tokat-Çorum region; Qc(f) ing to heat as the result of internal friction, then it could = (232±18)f(0.76±0.019) for Çorum-Adapazarı region; Qc(f) = be defined as intrinsic attenuation (Qi).The quality factor (280±28)f(0.79±0.021) for Adapazarı-Yalova region; Qc(f) = (Q) is a significant parameter that is used often in seis- (252±26)f(0.81±0.022) for Yalova- region. The mological studies such as seismic hazard, focus mecha- coda wave quality factor at all the lapse times and fre- nism solutions, calculation of magnitude and crust struc- quencies is Qc(f) = (206±15)f (0.85±0.012) in the study area. ture. There searchers use different techniques for deter- The most change of Qc with lapse time is determined at mining the attenuation of seismic waves. While these stud- Yalova-Saros region. The result may be related to hetero- ies are being carried out under various pressure and tem- geneity degree of rapidly decreases towards the deep crust perature conditions at laboratories, they are also able to like compared to the other sub region. Moreover, the high- be performed by using P, S or surface waves at site obser- est Qc is calculated between Adapazari – Yalova. It was vations [1, 20, 22, 23, 25, 30, 33]. The data obtained from interpreted as a result of seismic energy released by 1999 near or far field can be analyzed as per the scope ofthe Kocaeli . Besides, it couldn’t be established a study, and even the process can be performed by the time causal relationship between the regional variation of Qc or frequency domain. In recent studies researches have with frequency and lapse time associated to migration of used single scattering and multiple scattering models as the big earthquakes. These results have been interpreted suggested by [2] the most [14, 17, 24, 39, 40]. These mod- as the attenuation mechanism is affected by both regional els, where coda waves are being used, reveals the char- heterogeneity and consist of a single or multi strands of the acteristics of scattering that exist at the medium where is fault structure. located between the source and station. There is a strong relation between Qc and lapse time and frequency depen- dency. As the lapse time increases, information being ob- tained from depths of the earth increases. Thus, in cases where it has transferred to a more homogenous medium *Corresponding Author: Fadime Sertcelik: Kocaeli University –than the heterogeneity at upper parts- as it is preceded Kocaeli, Turkey; Email: [email protected] towards the depths, an increase in Qc is being observed. Mehmet Guleroglu: Email: [email protected]

Open Access. © 2017 F. Sertcelik and M. Guleroglu, published by De Gruyter Open. This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License Characteristics for North Anatolian Fault Zone Ë 481

Figure 1: (a) The earthquakes and fault rupture that had occurred in the recent century at North Anatolian Fault Zone. b) List of the earth- quakes that have occurred in the study area. NAFZ North Anatolian Fault Zone, DSFZ Dead Sea Fault Zone, EAFZ East Anatolian Fault Zone, NEAFZ North East Anatolian Fault Zone, GB Aegean Graben system.

The frequency dependency of Qc is an indicator forthe tec- a region in southern Germany. They calculated that the Q tonic complexity of the area. There have been attenuation is lower in the western NAFZ than in the eastern NAFZ for studies that carried out using various methods in some a distance ranging from 15 to 40 km. part of North Anatolian fault zone (NAFZ). [29] calculated The results of our study, investigating the attenuation that there are relatively low coda-Q values for 5 and 10Hz characteristic in detail of NAFZ have high tectonic activity at the Izmit - Sapanca and Iznik - Mekece fault segments level could provide a significant contributions. These are; in comparison with the north region of the fault zones. first, the crustal attenuation characteristichas been deter- [42] had found Qc values are to be very low at between 30 mined for the whole NAFZ area that is covering through and 200 at the Iznik - Mekece fault in the North Anatolian 12 provinces in which is included. Second, five Fault Zone and the frequencies from 2 to 10 Hz when com- sub regions were determined by considering where the pared with the values that range between 70 and 300 in large earthquake fractures are located. Qc change down the surrounding region. The reason for this is scattering to a depth of average 100 km had been determined with due to concentrated heterogeneous medium and temporal varying lapse time for each located region. Finally, the het- increase of intrinsic attenuation related to the preparation erogeneity degree of lower crust and upper mantle along for a large earthquake. [22] investigated the crustal struc- NAFZ has been determined. ture beneath the western part of the North Anatolian fault zone. The low-velocity and high-attenuationpatterns cor- relatefor most part with the fracturing zones of the NAFZ. 2 Tectonic Setting and Seismic While low velocities have been observed beneath the main sedimentary basins (e.g., Adapazarı, Düzce, and Kuzuluk), Activity there have been high-velocity and low-attenuation pat- terns correlated with blocks presumed to be rigid found North Anatolian Fault Zone (NAFZ), which was first named in (Kocaeli, Armutlu, and Almacik blocks). [15] has calcu- by [18], and is morphologically distinct having seismically lated the attenuation of coda wave at to active zones, which is among the most well-known faults beQc = 41f1.08 and Qs = 50f1.09 for 50 sec. lapse time. [5] in the world. The NAFZ is an active right-lateral strike-slip determined P wave attenuation of (Qp) along two portions fault with a length of 1200 km, and its width varies be- of the North Anatolian Fault Zone (NAFZ) in Turkey and in tween 100 m – 10 km. This fault constitutes as part of the 482 Ë F. Sertcelik and M. Guleroglu boundary in between Eurasia plate at north and Anatolian by 1.5 or 2 corresponding to the difference from theorigin plate at South. It extends from Karliova at east to Saros time of earthquake and the arrival time of the S wave.The Bay at west and is parallel to the shores of the , reason for this is to prevent the coda wave from being af- and connects the East Anatolian compression area to the fected by the source [34]. Derived coda Q values show a Aegean – Cyprus arch [6–8] (Figure 1). clear dependence on frequency according to the relation- n In the recent century, the earthquakes that have oc- ship Q = Qof , where Q0 is the value of Qc at 1 Hz, and n rep- curred on NAFZ fault indicate that the earthquakes have resents the degree of frequency dependence of Qc. n value systematically migrated from east to west (Figure 1-a).The is effected from the heterogeneity of structure and alsoQo migration to west, was started in 1939 by the Erzincan has low values (<200) in the regions where the tectonic ac- earthquake that was Ms = 8.2, and concluded with the tivity is high. Izmit earthquake (Mw = 7.4) on August 17, 1999, and Duzce [32], in its Qc estimation was made by using coda earthquake (Mw = 7.2) on November 12, 1999. It has been wave, that referto the identification of a surface in the form emphasized by many studies that have been performed, of an ellipsoid. The projection of this ellipsoid on the sur- there is an increased chance of a major earthquake occur- face source is defined as a =Vst/2 for large axis and b = 2 2 1/2 ring on the strand of the NAFZ in the (Fig- a -(r /4) for small axis. Vs is considered to be a S wave ure 1-b) [19, 31, 41] and [26]. at a velocity of 3.7 km/h, r as distance between source and receiver, and t as the lapse in time. And the maximum depth of this ellipsoid volume which varies depending on the distance between focus and station, is calculated by h 3 Coda and Single Scattering 2 2 1/2 = [(vt/2) -(r/2) ] + hav. Where h is the maximum depth; Method hav is the average hypocenter depth of the earthquakes; r is hypo central distance. In this study, the maximum depths The coda waves are secondary waves that compose of scat- varied between 89 to 98 km (Table 1). tering body waves due to the heterogeneity in the crust and mantle. As lapse time increases, the coda waves that arise as the result of single (<100s) or multiple (>100s) scattering 4 Data are analyzed ([10, 21]). As the S wave usually covers the P wave especially in near the surface earthquakes, this phe- In this study, 654 earthquakes with a magnitude of M≥ nomena leads to S coda wave generally being used instead 3.0 which had occurred between the years of 2009 to 2014 of P coda in such studies. in North Anatolian Fault Zone were used. The stations be- In this study, single scattering method –which was ing used in this study are operated by KRDEA, and gener- first used by [2] has been used. In this method, the ampli- ally, the recorders hada three component broadband. The tude decrease in coda wave is being defined as a function study area has been divided into five areas based onthe of time. fault fracture where large earthquakes had occurred, the A(f , t) = C(f)t−α exp(−πŠ/Qc) (1) distribution of earthquake and the stations. The distribu- tion of earthquakes and stations used in each region can In equation (1), A(f,t) is the (f) filtered coda wave amplitude be seen in Figure 2a. at a specific central frequency; t is the lapse time; Qc is the I.Region situated between Karliova - Tokat which ap- quality factor of coda wave; and C = 2(2∆f)1/2. Moreover, α proximately covers the fault of the 1939 Erzincan earth- is 1, 0.50 or 0.75 for body, surface and scattering waves [36]. quake (M = 7.9). Erzincan / Cayirli earthquake 1939 (M = For this reason, if logarithm is calculated by having α = 1 5.9),1942 Tokat / Erbaa and Niksar earthquakes and 1992 in the equation (1), it becomes; Erzincan / Cayirli earthquake (M = 6.8) occurred in this ln(A(f , t)) = ln C(f ) − πŠ/Qc(f) (2) area. Karliova triple junction also located in the region. For calculation Qc of the region were analyzed 280 seismo- For each selected lapse time window and frequency, the grams of 274 earthquakes.The magnitudes of earthquakes Qc is calculated from the inclination of the line passed varied between 3.0≤ M≤ 5.6, and their depths were 2.8 to through the scattered values in ln(A(f,t)*t) and t graph. In 12.9 km. The records of BAYT, ERZN, ILIC, PTK, RSDY sta- the calculation of Qc(f), the RMS magnitudes of coda enve- tions were used. lope at a lapse time window is used. The lapse window is II.Region located between Tokat - Corum which ap- selected from the coda of S wave. According to various re- proximately covers the fault segment of Samsun / Ladik searchers, the initial time of the window should be product Characteristics for North Anatolian Fault Zone Ë 483

Figure 2: a) Epicentral location of earthquakes used in the study (white circles), location of station (red triangles) and of 5 sub regions (black ellipse). Qo and n maps for a lapse time of 20 sec. at I, II, III, IV, V sub region, (b1, b2, b3, b4, b5) and (c1, c2, c3, c4, c5), respectively. earthquake (M = 7.4) in 1943. A number of earthquakes quake in 1999. 198 records of 61 earthquakes were used. with a magnitude of over 3.0 at Amasya and Corum have The magnitude of earthquakes is between 3.0M4.3, and occurred in the last 5 years showing that the area is ac- their depth wasbetween 2.7-13.8 km. In the analysis of the tive. 154 records of 141 earthquakes were recorded byBZK, region, the ADVT, ARMT, CAVI, GEMT, GPA, GULT, HRTX, CANT, CORM, DIKM, KVT, TOS, YOZ stations were used. ISK, KLYT, MDNY, SILT, SPNC stations were used. The magnitudes of earthquakes were between 3.0≤ M≤ 4.5, V. Region is the one between Yalova – Saros Bay and their depths were 3.5 to 10 km. which approximately covers the fault fracture of the earth- III. Region approximately covers the fault of earth- quake of 1912. From the area 214 records of 81 earthquakes quakes that occurred between the years 1944 to1951 in the –whose magnitudes varied between 3.0≤ M≤ 5.0, and area between Corum - Adapazari. 1944 Karabuk / Safran- whose depths ranges from 5 to 18 km- had been used.The bolu earthquake (M = 7.2), 1951 Cankiri / Cerkes earthquake data from EDC, ISK, KCTX, ERIK, EZN, GADA, GELI, LAP, (M = 6.9) and 1957 Duzce / Golyaka earthquake (M = 7.1) oc- KRBG, ENEZ, MDNY, RKY stations was used. curred in the region. The biggest recent earthquake in this In this study, seismograms that have signal/noise ra- area is the Duzce / Kaynasli earthquake and it occurred tio greater than 2 were selected.The length of the time win- on 11.12.1999 (Mw = 7.2).132 records of 97 earthquakes – dows is related to the change of coda Q. In this study, the which were recorded by CANT, CORM, GULT, KDZE, LOD, length of the lapse time window was taken at 20-30-40 sec MDUB, TOS, YOZ stations- were analyzed. The magnitudes respectively. Afterwards it was filtered by 6 units of con- of earthquakes vary between 3.0≤ M≤ 4.8, and their depths secutive Butterworth filters with 8 poles whose frequencies range from 2.6 to 11.5 km. were 1.5, 3, 6, 9, 12 and 18 Hz, and whose bandwidths were IV. Region is the one between Adapazari - Yalova 1, 2, 4, 8, 12 and 24 Hz respectively. Moreover, envelopes which approximately covers the fault fracture of the earth- of S wave coda at local earthquakes had differed slightly 484 Ë F. Sertcelik and M. Guleroglu

Table 1: Table Qc values using –obtained by using single scattering model- and the standard deviation. N indicates the number of used earthquakes for each subregion and along the fault.

Central Frequency (Hz) Region I Region II Region III Region IV Region V All Region (Karliova- (Tokat- (Çorum- (Adapazari- (Yalova- (NAFZ) Tokat) Çorum) Adapazari) Yalova) Saros Gulf) Eartquakes Nums (N) 274 141 97 61 81 654 Max Depth (km) 92.5 89.6 91.0 93.0 98.0 98.0 Lapse Time 20 sec 1.5 228(±15) 237(±17) 260(±11) 300(±25) 269(±22) 262(±19) 3 416(±16) 416(±28) 440(±31) 545(±27) 494(±19) 489(±22) 6 761(±21) 730(±23) 745(±27) 990(±32) 910(±35) 913(±33) 9 1082(±52) 1014(±45) 1015(±47) 1403(±51) 1300(±50) 1315(±49) 12 1390(±63) 1280(±54) 1262(±61) 1797(±60) 1674(±73) 1703(±68) 18 1978(±69) 1777(±61) 1718(±77) 2546(±71) 2392(±76) 2454(±81) Lapse Time 30 sec 1.5 274(±12) 312(±10) 316(±18) 395(±16) 355(±14) 333(±17) 3 494(±32) 521(±25) 542(±28) 684(±25) 619(±28) 596(±26) 6 890(±32) 870(±30) 930(±43) 1182(±45) 1078(±42) 1068(±39) 9 1256(±54) 1174(±49) 1277(±53) 1628(±69) 1490(±53) 1501(±56) 12 1604(±76) 1453(±65) 1598(±43) 2044(±88) 1876(±63) 1911(±68) 18 2264(±84) 1961(±79) 2192(±86) 2815(±76) 2595(±65) 2686(±101) Lapse Time 40 sec 1.5 309(±28) 347(±23) 382(±21) 487(±15) 445(±18) 393(±21) 3 557(±30) 575(±38) 638(±31) 807(±48) 744(±22) 689(±18) 6 1004(±53) 954(±51) 1066(±48) 1339(±54) 1243(±45) 1208(±52) 9 1418(±62) 1283(±55) 1439(±51) 1800(±72) 1677(±56) 1678(±65) 12 1810(±84) 1583(±79) 1780(±76) 2221(±95) 2075(±84) 2118(±84) 18 2555(±110) 2128(±85) 2403(±95) 2986(±99) 2802(±122) 2941(±109) at horizontal and vertical components [36]. In this study, The graphs for the frequency dependent Qc change at single scattering method was applied to vertical compo- varying lapse times for each area is being shown in Fig- nent data. Coda window is started from the time that corre- ure 4. The number of earthquakes (N), Qc values calcu- sponds to twice times of the differences between the origin lated at different lapse times and frequencies and standard time and the S wave arrival time. deviations was shown in Table 1. The distribution maps In the equation (1), different Qc values for the different of Qo and n values – which were calculated for each sub lapse timeswas calculated by using the least squares re- region and lapse time of 20 sec- is depicted in Figure 2 gression method. In the calculations used the data which b1,b2,b3,b4,b5 and c1, c2, c3, c4, c5, respectively. had a correlation coefficient higher than 0.75 was used. In Figure 3, the BAYT station’s vertical component record of the earthquake -which occurred on 02.25.2010, at 10:06:25- 5 Result and Discussion is being shown. In this example, the coda window length is 20 sec. The left six panels show the seismogram that was Frequency dependent coda Q analysis at different lapse filtered at the central frequencies that were mentioned ear- times reveals the average attenuation characteristic of lier. On the right side of these six panels, is the Qc that can each sub region within the study area. be obtained from the inclination of the line that is fitted At I. Region: Qc(f) = (189±26)f(0.86±0.02) was calcu- with least squares method. The correlation coefficient val- lated in the region.When compared to all regions, the ues are higher than 0.76. calculated Qo in all lapse time is the lowest, and fre- Characteristics for North Anatolian Fault Zone Ë 485

Figure 3: Unfiltered (top) Z component signal of earthquake recorded by BAYT station at Region I on 02.25.2010 at 10:06:25. 20scodawin- dows used in Qc calculus. Signal was filtered in central frequency at 1.5, 3, 6, 9, 12, 18 Hz (left). The straight line is fitted inaleastsquare method. The estimated Qc value for each frequency component is also shown (right) 486 Ë F. Sertcelik and M. Guleroglu

Figure 4: Plot of average values of Qc with different central frequencies at 20-30-40 sec. lapse time windows in five subregions quency dependency (n) is the highest in the subregion uation study for both the S wave and coda wave for 20- (Figure 2.b1-c1). And the highest Qo value was calculated 50 sec. lapse timesand at 1.5-24 Hz frequencies at Erzin- around Cayirli and Uzumlu in the area. Also, the n val- can area. They had especially specified that a decrease ofS ues (0.8-1.1) that indicates the frequency dependency in- wave and coda wave is very different at high frequencies. creases from east to west in the area. [4] performed atten- Characteristics for North Anatolian Fault Zone Ë 487

This difference is not occurred when it happens at short distance and low frequencies. Region II indicate decreasing values for 1/Qo and n. Qo values 171, 231,258 and n values 0.81, 0.74, 0.73 were calculate for increasing lapse times.The values of Qc were obtained a regional regression relation of form Qc(f) = (216±19)f(0.76±0.018) for Tokat-Çorum region.The variation of Qo and n values for 20sec lapse time wereshown in Fig- ure 2.b2-c2. Around Mecitozu –where an earthquake oc- curred in 1997 (Md = 6.0), shows that the Qo had the lowest value and n had the highest value of the area. The degree of heterogeneity in the upper crust slowly changes to ho- Figure 5: Comparison of the mean values of Qc as a function of fre- mogeneity towards the depths. quency obtained at three lapse time windows in all sub regions In the Region III, Attenuation frequency dependency function for the region was(232±18)f(0.76±0.019).Qo = 191, preted as this area having a fast decrease in heterogene- 203,283 and n = 0.76, 0.78, 0,74 were calculated for 20, 30, ity along the depth (Figure 4). Especially, as Qc result for 40 sec lapse times (Figure 4).The low Qo and high n values 20sec lapse time are showing that Tekirdag basin is the for 20 sec lapse timein the between Bolu- Çankırı reveal it location where the lowest of heterogeneity and tectonic to be a very heterogenic structure at shallow depths (Fig- activity. (Figure 2.b5-c5). However, there is a relative in- ure 2.b3-c3). This occurrence is supported by the hot spring crease in frequency dependency towards Saros Bay. For in the region, as they indicate a presence of deep-melted that area, the S wave quality factor has been calculated as particles. [24], in their study performed at northwest Hi- being 22±3f 1.15±0.09 by [16]. malayas, determined that a hot substance in a deep can Looking along the fault zone, the Qo is varying in cause high attenuation. between 136-307, n has values between 0.72-0.96 at in- Region IV is in the north branch of NAFZ, and this creasing lapse time from 20 to 40 sec (Figure 5). Qo val- branch enters in the graben of Izmit – Sapanca, and dis- ues are shown as decreasing from Corum to Karliova appears at the south of Ganos and at west shore of Mar- where located at the junction of the North Anatolian Fault mara Sea. Furthemore,the Kocaeli / Golcuk earthquake (M Zone (NAFZ) and the East Anatolian Fault Zone (EAFZ) = 7.4) of August 17, 1999occurred on this branch. An av- (Figure 6). Average frequency-dependent relationship has erage frequency dependent power law fit for the region been estimated for Qc(f) as (206±15)f (0.85±0.012). Despite may be given as Qc(f) = (280±28)f(0.79±0.021). While an in- this there is not much change from Corum to Saros Gulf, crease in Qo value has been observed at the north branch as there is a small increase in Qo values at Saros Gulf, of the fault, a decrease in the n value is also occurring (Fig- eastern Marmara Region, and Bolu. While the n value is- ure 2.b4-c4). The average Qo value (212-362) and n value the highest in the area between Bolu and Çankırı, it has (0.73-0.86) were calculated for increasing lapse time in the the lowest value in Yalova, Bolu, Karlıova and around Ço- area (Table 1, Figure 4). The highest Qc values being calcu- rum, even with the n value is increasing from Karlıova to lated in this area. It is seen thatstress has decreasedwith Tokat and from Yalova to Soros Gulf. This is stillinterpreted 1999 Kocaeli earthquake. And NAFZ there have been val- as resulting from the variationstectonic structure and het- ues found that correspond to this being an active area. In erogeneity among the different areas. The change in Qc the S wave attenuation study performed at [16] for the same as the depth increases, in other words as the lapse time area, they found Qs = (13±1)f(1.22±0.05). increases has become a subject which many researchers Region V includes the NAFZ –that proceeds intothe have become interested in.Such studies are numerous es- Marmara Sea- and Saros Bay. The Marmara Sea area is the pecially inactive tectonic areas. In the studies when the transition zone between the strike-slip deformation zone lapse timeis 20 sec., [9] calculated Qc = 29f0.9 in their study and expansion area and west Anatolian [8]. The quality at Etna and Granada, [12] calculated Qc = 64f0.9in their factor of coda waves (Qc) estimated by back scattering study at volcanic area in the southeastern area of Sicily, model show strong frequency dependence and is calcu- [13] calculated Qc = 66f1.16 in the Koyna area, and in the lated as (252±26)f(0.81±0.022) .As the lapse time increase, studies performed with a lapse time of 20 sec., Qc = 96f1.09 Qo value changes from 188 to 330, and n value changes had been calculated, and in the studies performed with a from 0.74 to 0.88. This area shows the highest change lapse time 40 sec. Qc = 131f1.04 had been calculated. [28], within the lapse time along the fault, this has been inter- 488 Ë F. Sertcelik and M. Guleroglu

Figure 6: The maps of frequency dependent (n) (bottom) and coda Q distribution at 1Hz (Qo) (top) in North Anatolian Fault Zone in their study performed in the northwest Himalayas in The result of the study performed, Qc(f) = (206±15)f 2006, they obtained the result of Qc = 113f1.01 for a lapse (0.85±0.012) was calculated in average along the fault at all time of 30 sec., and of Qc = 153f1.07 for a lapse time of 40 data and lapse times. The average Qc value for NAFZ is sec. And in the studies performed in Turkey, [3] had calcu- varying from 262(±19) at 1.5 Hz to 2454(±81) at 18 Hz for lated Qc = 50.7f1.01 for a lapse time of 30 sec. in their study 20 sec lapse time, while it is varying from 393(±21) at 1.5 Hz performed in western , and [35] had calculated to 2941(±109) at 18 Hz for 40 sec lapse time. For the whole the coda wave attenuation to be Qc = 95f0.96 in the south- fault, the variations in Qc with lapse time are calculated west Anatolia, and [38] had calculated Qc = 45f0.76 for 30 as, sec. andcoda wave analysis was also performed at Karliova Qc(f) = (136 ± 25)f (0.96±0.027) at lapse time 20sec triple junction. [37] had calculated the coda wave attenu- Qc(f) = (208 ± 22)f (0.85±0.02) at lapse time 30sec ation at North Anatolian Fault Zone to be Qc = 37f0.90 for Qc(f) = (307 ± 28)f (0.72±0.025) at lapse time 40sec. 20 sec. lapse time, as Qc = 60f0.82 for 30 seclapse time and Low Qcvalue indicates the presence of active and het- as Qc = 85.8f0.75 for 40 sec lapse time. [27], in their study erogeneous structure along the fault. performed in Israel, calculated Qc = 77f0.96 for the north- The frequency dependent Qc value is calculated as, ern area, and Qc = 126f1.05 for the whole Israel. [11], in their Qc(f) = (189 ± 26)f (0.86±0.02) study performed in Faryab area, calculated Qc = 28f1.14 for R2 = 0.97 for Karlıova-Tokat region (Region I) 5 sec. lapse time, and Qc = 105f0.98 for 30 sec lapse time. Qc(f) = (216 ± 19)f (0.76±0.018) R2 = 0.96 for Tokat-Çorum region (Region II) Qc(f) = (232 ± 18)f (0.76±0.019) 2 6 Conclusion R = 0.96 for Çorum-Adapazarı region (Region III) Qc(f) = (280 ± 28)f (0.79±0.021) R2 = 0.95 for Adapazarı-Yalova region (Region IV) This study is the first comprehensive research in which the Qc(f) = (252 ± 26)f (0.81±0.022) coda wave attenuation feature was determined along the R2 = 0.96 for Yalova- Gulf of Saros region (Region V) North Anatolian Fault Zone at 20, 30, 40 sec. lapse times for five sub regions. The lowest Qo and highest frequency and 1.5, 3, 6, 8, 12 and 18 Hz center frequencies were used. dependent (n) values were in Region I and have been in- terpreted to result from the Karliova triple junction. The Characteristics for North Anatolian Fault Zone Ë 489 highest Qc is calculated between Adapazari – Yalova. It is the vicinity of the Iznik-Mekece fault, the North Anatolian Fault explained as a result of seismic energy released by 1999 Zone, Turkey, Bull. Earthq. Res.1988, 63, 327–348. Kocaeli Earthquake.The values of Qc with 20, 30 and 40sec [12] Rautian G., Khalturin I.,The use of the coda for determination of the earthquake source spectrum, Bulletin of the Seismological lapse time shows regional change along the fault zone. The Society of America, 1978, 680, 923-948. rapid change of Qc was found between Çorum–Adapazarı, [13] Sahin S., Lateral variations of coda Q and attenuation of seismic and the smallest change ofQc occurred betweenYalova – waves in Southwest Anatolia, Journal of Seismology, 2008, 12, Saros. It is seen that there are regional differences a transi- 367–376. tion to homogeneity to low crust and upper mantle. Also, [14] Del Pezzo E., Patanè D., Coda Q dependence on time, frequency the changes from region to region in Qc were explicit as and coda duration interval at Mt. Etna, Sicily, Editors: Gasparini P., Scarpa R., Aki K., Volcanic Seismology, 3nd ed, Springer- they were affected by the structure of fault rather than the Verlag, New York, 1992, 109–119. ordering of earthquake migrating from east to west. Addi- [15] Horasan G., Boztepe, A., S-wave attenuation in the Sea of Mar- tionally it has been determined that the frequency depen- mara, Turkey, Phys. Earth Planet. Interiors, 2004,142, 215–224. dency is high and Qo value is low in areas where the NAFZ [16] Pulli J.J., Attenuation of coda waves in New England, Bull. Seis. lies along a single branch. Soc. Amer., 1984, 74, 1149–1166. [17] Gao L., Lee L., Biswas N., Aki K., Comparison of the effects be- tween single and multiple scattering on coda waves for local Acknowledgement: We are thankful to Bogazici Univer- earthquakes. Bulletin of the Seismological Society of America, sity Kandilli Observatory National Earthquake Monitoring 1983, 73, 377-389. Center from where the data used in this study is obtained. [18] Ketin İ., Kuzey Anadolu Deprem Fayı, İTÜ Dergisi, 1957, 15, 49-52 [19] Koulakov I., Bindi D., Parolai S., Grosser H., Milkereit C., Distri- bution of seismic velocities and attenuation in the crust beneath the north Anatolian Fault (Turkey) from local earthquake tomog- References raphy, Bull. Seism. Soc. Am., 2010,100(1), 207–224 [20] Aki K., Chouet B., Origin of Coda waves: Source, attenuation and scattering effects, Journal of Geophysical Research, 1975, 80, [1] Aki K., Analysis of the seismic coda of local earthquakes as scat- 3322-3342. tered waves, Journal of Geophysical Research, 1969, 74, 615- [21] Meirova T., Pinsky V., Seismic wave attenuation in Israel region 631. estimated from the multiple lapse time window analysis and S- [2] Bozkurt E., Neotectonics of Turkey-a synthesis, Geodinamica wave coda decay rate, Geophyscial Journal International, 2014, Acta, 2001, 14, 3-30. 197, 581-590. [3] Sertçelik F., Estimation of coda wave attenuation in the East [22] Akinci A., Taktak G., Ergintav S., Attenuation of coda waves in Anatolia Fault Zone Turkey, Pure and Applied Geophysics, 2012, Western Anatolia, Physcial Earth Planet Interiors, 1994, 87, 155– 169, 1189–1204. 165. [4] Parsons T., Toda S., Stein R.S., Barka A., Dieterich J.H., Height- [23] 23 Akinci A., Eyidogan H., Frequency-dependent attenuation of ened odds of large earthquakes near Istanbul: an interaction- S and coda waves in Erzincan region (Turkey) Physics of the based probability calculation. Science 2000. Earth and Planetary Interiors, 1996, 97, 1, 109–119. [5] Jin A, Aki K., Spatial and temporal correlation between coda q [24] Gholamzadeha A., Rahimi H., Yaminifard F., Spatial and Tem- and seismicity in china. Bull Seismol Soc Am., 1988, 78(2):741– poral Variation of Coda-Wave Attenuation in the Faryab Region, 769 Southeast of the Sanandaj–Sirjan Zone, Using Aftershocks of [6] King G.C.P.,Hubert-Ferrari A., Nalbant, S.S., Meyer B., Armijo R., the Tiab Earthquake of 28 February 2006, Bulletin of the Seis- Bowman D., Coulomb interactions and the 17 August 1999 Izmit, mological Society of America, 2014, 104, 529-539. Turkey earthquake. Earth Planet. Sci., 2001, 333, 557–569. [25] Akinci A., Mejia J., Jemberie A.L., Attenuative dispersion of P [7] Kita S., Nakajima J., Hasegawa A., Okada T., Katsumata K.,Asano waves and crustal Q in Turkey and Germany. Pure and Applied Y and Kimura T.,Detailed seismic attenuation structure beneath Geophysics, 2004, 161(1), 73-91. Hokkaido, northeastern Japan: Arc-arc collision process, arc [26] Kuteynikova M., Tisato N., Jänicke R., Quintal B., Numerical magmatism, and seismotectonics Journal of Geophysical Re- modeling and laboratory measurements of seismic attenuation search: Solid Earth, 2014, 119, 8,6486–6511. in partially saturated rock, 2014, doi: 10.1190/geo2013-0020.1, [8] Kopnichev Y.F., The role of multiple scattering in the formation 79, 2, L13-L20. of seismiogram’s tail. Izv. Acad. Sci. USSR, Phys. Solid Earth, [27] Stein R. S., Dieterich J. H., Barka A. A., Progressive failure on the 1977, 13, 394-398. North Anatolian fault since 1939 by earthquake stress trigger- [9] Rachman A. N., Woong Chung, T., Yoshimoto K., and Son B., ing. Geophysical Journal International, 1997, 128, 3, 594-604. Separation of Intrinsic and Scattering Attenuation Using Single [28] 28 Sato H., Fehler M., Seismic Wave Propagation and Scatter- Event Source in South Korea Bulletin of the Seismological Soci- ing in the Heterogeneous Earth, Springer and Verlag, 308, New ety of America, 2015, doi:10.1785/0120140259. York, 1998.288, 661–665 [10] Le Pichon X., The North Anatolian Fault in the Sea of Marmara, [29] 29 Havskov J., Malone S., McClurg D., Crosson R., Coda q for the J. Geophys. Res., 2003, 108(B4), 2179. state of Washington. Bull Seismol. Soc Am 1989, 79(4),1024– [11] Tsukuda T., Satake K., Honkura Y., Uçer S. B., Isikara, A. M., Low 1038. seismicity, low coda-Q and discontinuities of the upper crust in 490 Ë F. Sertcelik and M. Guleroglu

[30] Ambraseys N.N., Some characteristic features of the Anatolian [37] Singh C., Mondal P., Singh S., Mohanty D.D., Jaiswal N., Ku- fault zone, Techtonophysics, 1970,9, 143–165. mar M.R., Lg attenuation tomographic models of Himalaya and [31] Kumar A., Sinvhal A., Joshi A., Kumar D., Kumar S.P., Coda wave southern Tibet. Tectonophysics, 2015, 664, 176–181. attenuation characteristics for Kumaon and Garhwal Himalaya, [38] 38 Singh C., Bharathi V.S., Chadha R., Lapse time and India, 2014,75, 2,1057-1074. frequency-dependent attenuation characteristics of Kumaun hi- [32] Pandit P., Kumar D., Muralimohan T.R., Niyogi K., Das SK., Esti- malaya. Journal of Asian Earth Sciences 2012, 54–55, 64–71. mation of seismic Q using a non-linear (Gauss-Newton) regres- [39] Giampiccola E., Gresta S., Rascona F., Intrinsic and scattering sion. Geohorizons, 2011, 19–23. attenuation from observed seismic codas in southeastern Sicily [33] Barka A.A., The north Anatolian fault zone. In Annales tectoni- (Italy), Physics Earth Planet Interior., 2004, 145, 55–66 cae, 1992, 6, 164-195 [40] Gupta C., Teotia S., Rai S., Gautam N., Coda Q estimates in the [34] Nishigami K.Y., Ito Y., Gurbuz C., Pınar A., Aybey N., Ucer S.B., Koyna region, Pure Application Geophysics, 1998, 153,713–731. Honkura Y., Isıkara A.M., Kurenaı, Kyoto University Research In- [41] Kumar S., Singh P., Singh P., BiswalS., Parija MP., Frequency de- formation Repository. Bulletin of the Disaster Prevention Re- pendent attenuation characteristics of coda waves in the North- search Institute. 1990, 40(2), 41-56. western Himalayan (India) region, Journal of Asian Earth Sci- [35] Sertcelik F., Sertçelik, I., Karlıova üçlü ekleminde koda dalga ences, 2016,17, 337–345. analizi, UygulamalıYerbilimleri Dergisi, 2008,2, 31-37. [42] Horasan G., Kaşlılar A., Boztepe A., Türkelli N., S- wave atten- [36] Mukhopadhyay S., Tyagi C., Rai S., The attenuation mechanism uation in the Marmara Region, Geophysical Research Letters, of seismic waves in northwestern Himalayas, Geophyscial Jour- 1998, 25, 2733-2736. nal International, 2006, 167, 354–360.