Variation of Coda wave attenuation in two different Tectonic Areas in the Iranian plateau, Alborz and NW Zagros

The attenuation of coda wave in Alborz and NW Zagros has been estimated using the single backscattering (SBS) and the single isotropic scattering (SIS) methods at 9 frequency bands with central frequencies of 1, 1.5, 2, 3, 4, 6, 8, 12 and 16 Hz. The database consisted of 677 local earthquakes (418 events in the Alborz region and 259 events in the NW Zagros region) with the MN magnitude range from 3.0 to 5.7. A total of 8,717 seismograms with signal-to-noise ratios of greater than 3 and epicentral distances less than 200 km were used. To investigate the depth variation of attenuation in this study, the coda quality factor Qc was estimated in each area at epicentral distance range of R<100 km and 100<R<200 km, through 11 coda window lengths between 10 and 60 s. The estimated average frequency-dependent relation of Qc = Q0f n at coda window length of 10 to 60 s varies from 66 f 1.04 to 164 f 0.72 in Alborz and 66f 0.99 to 157 f 0.76 in NW Zagros. In this study, the value of the frequency-dependent parameter decreases with increasing the coda window lengths, and suggests that the lithosphere becomes more homogenous with increasing depth. Furthermore, the coda wave quality factor was determined from all three components.We found that there is no significant difference in obtained results by using the vertical and horizontal components. This indicates that the seismic waves encounter similar heterogeneities and attenuation in the vertical and horizontal directions in the study area.The average frequencydependent relations of coda waves have been derived Qc=109f 0.87and Qc= 108f 0.86, for the Alborz and NW Zagros regions, respectively, usingthe SBS model at the coda window length of 25s. Comparison of estimated coda-Q values show that, the obtained Qc values using the SIS method are slightly higher than that those obtained using the SBS method.


Introduction
The seismic attenuation, as an important characteristic in the modern seismology, has an inevitable effect on the determination of the source mechanism, the waveform modeling, the simulation of strong ground motion and the seismic hazard analysis [Jackson and Anderson, 1974;Farrokhi et al., 2016;Soham and Abhishek, 2016].
The attenuation of seismic waves in the lithosphere is also an important property for studying the regional earth structure and seismotectonic activity [Kumar et al., 2005].
Seismic attenuation as amplitude decay through wave propagation is caused by two distinct physical processes: elastic and anelastic properties of the medium. Attenuation is usually expressed as the inverse of the quality factor (Q) and is generated by two main sources: scattering due to heterogeneities and intrinsic absorption. The existing random heterogeneities throughout the earth's lithosphere are responsible for the generation of late-arriving wave trains in the tail portion of seismograms -after the arrival of major wave types such as P, S, and surface wavesrecords of the local and regional events, which are called 'coda waves' [Aki, 1969;Aki and Chouet, 1975;Sato, 1977;Parvez et al., 2008;Sato et al., 2012;Obermann et al., 2012;Havskov et al., 2016]. Coda waves are backscattered body waves [Aki, 1969;1980], which their amplitude decreases due to attenuation (including scattering) and geometrical spreading [Havskov et al., 2016]. Aki and Chouet [1975] and Sato [1977] proposed the single scattering model to calculate the attenuation of seismic waves from the decay of coda waves. The multiple-scattering model which is suggested by Gao et al., [1983] is another model to estimate the attenuation of coda waves [Kumar et al., 2005]. Unlike the earlier part of the coda waves, the effect of secondary and tertiary scattered waves is not negligible for the later part [Kopnichev, 1977], therefore, the multiple-scattering model is mainly for lapse times greater than 100 s [Gao et al., 1983;Kumar et al., 2005;Sedaghati and Pezeshk, 2016a]. At very long lapse times, the coda-Q is almost entirely influenced by intrinsic absorption [Shapiro et al., 2000;Hovskov et al., 2016]. In other words, at long lapse time, coda waves enter in the diffusive regime, which implies Q c -1 =Q i -1 in a simple uniform half-space [Shapiro et al., 2000]. It should be pointed out that the estimated coda-Q from the single scattering model represents an effective attenuation including intrinsic absorption and scattering loss [Shapiro et al., 2000;Jin and Aki, 2005]. In recent years, the importance of multiple-scattering model has grown and many studies have been carried out by this model [e.g., De Siena et al., 2016;Galluzzo et al., 2015;Havskov et al., 2016].
Nonetheless, the single scattering model still remains a well-accepted model to comprehend the tectonic processes at the crustal and lithospheric scale, because it offers an easier way to estimate the attenuation properties of the Earth [Irandoust et al., 2015].
After the advent of the coda wave theory by Aki, [1969], the attenuation of coda waves has been estimated for different regions in the world using single-scattering model [e.g., Singh and Hermann, 1983;Mak et al., 2004;Ma'hood and Hamzehloo, 2009;Rahimi et al., 2010;Sertçelik, 2011;Havskov et al., 2016;Sedaghati and Pezeshk, 2016a]. Aki, [1969] as a pioneer of the single scattering method, used the local earthquakes with a magnitude range of 2.5-5.0 to analyze the seismic coda in the Parkfield region. By applying this method, numerous researchers [e.g., Aki, 1969;Kumar et al., 2005;Kim et al., 2006;Chung et al., 2009;Rahimi et al., 2010;Sharma et al., 2015;Ma'hood, 2014;Farrokhi et al., 2015;Sedaghati and Pezeshk, 2016a;Dobrynina et al., 2017] have estimated the Q c values using the local earthquakes with different magnitude ranges of 0.6-6.1. The successful application of the single scattering method for estimating Q c value in active/stable regions using local earthquakes with M<6.0, has been proven [Irandoust et al., 2015].
As mentioned earlier, the inverse of the quality factor (Q), which indicates the wave transmission quality of the bedrock, represents the attenuation properties of the medium [Chandler et al., 2006]. Generally, the quality factor depends on the frequency of the seismic waves, therefore, it is expressed by the power-law equation as follows: Q=Q 0 f n , where f is frequency, Q 0 and n are constants that represent the quality factor at 1 Hz and frequencydependent parameter, respectively. These constants vary from region to region according to the attenuation characteristics of the seismic waves [Aki and Chouet, 1975;Singh and Hermman, 1983]. Previous studies of the seismic quality factor in different regions in the world have shown that low Q 0 values (Q 0 <200) with higher n values (n>0.7) are indicators of tectonically active regions [Aki and Chouet, 1975;Roecker et al., 1982;Sato and Fehler, 1988;Mak et al., 2004;De Lorenzo et al., 2013;Ma'hood, 2014;Sedaghati and Pezeshk, 2016a]. High values of Q 0 (Q 0 >600) with low n (n<0.4) represent tectonically inactive and stable regions [Singh and Herrmann, 1983;Pulli, 1984;Pujades et al., 1991;Sedaghati and Pezeshk, 2016a]. Moreover, some studies reported intermediate values for Q 0 and n (200<Q 0 <60, 0.4<n<0.7) in tectonically moderate regions [Scherbaum and Kisslinger, 1985;Kumar et al., 2005, Sedaghati andPezeshk, 2016a].
Various studies have been carried out to investigate the seismic wave attenuation in the Alborz and Zagros regions using different datasets and methods [e.g., Motazedian, 2006;Rahimi et al., 2010;Motaghi and Ghods, 2012;Naghavi et al., 2012, Farrokhi et al., 2015. In this study, first, the attenuation of coda waves is estimated in the Reza Emami et al. Fitting elevation and geoid anomaly data combined with thermal analysis.
Alborz and NW Zagros regions using the Single Backscattering (SBS) [Aki and Chouet, 1975] and Single Isotropic Scattering (SIS) [Sato, 1977] models. Then, obtained Q c values using both vertical and horizontal components in the Alborz and NW Zagros regions are compared.
Many researchers [e.g., Rautian and Khalturin, 1978;Roecker et al., 1982], have been shown that the estimated Q-coda values at long lapse times (large epicentral distances) are greater than those for short lapse times (short epicentral distances). The dependence of Q c values on the coda window length and its variation with depth is another topic of interest that has been investigated in this study. Finally, our results are discussed and compared to other studies across the world and in Iran.

Geology and tectonic setting of the study area
The Iranian plateau is a relatively wide zone of compressional deformation along the Alpine-Himalayan active mountain belt, which is entrapped between two stable platforms, the Arabian plate in the southwest and the Turan platform (Eurasia) in the northeast [Mirzaei et al., 1998;Rahimi and Hamzehloo, 2008]

Alborz
The Alborz Mountains are located in northern Iran, parallel to the southern margin of the Caspian Sea. These Mountains are a still active collisional belt with 3-5 km topography, a length of 600 km in roughly E-W direction and a width of 100 km in N-S direction. They are surrounded by the Talesh Mountains in the west, Binalud and the Kopet-Dagh mountains in the east, the South Caspian Basin in the north and central Iran in the south. Alborz is characterized by the dominance of platform-type sediments, including limestone, dolostone, and clastic rocks. Rock units from Precambrian to Quaternary have been identified, with some hiatuses and unconformities on Paleozoic and Mesozoic.
Mesozoic sediments with a thickness of up to 3 km and 1-2 km of Cenozoic, mostly synorogenic sediments and volcanic activity throughout the Cenozoic, starting in Late Cretaceous, have covered its basement which includes Paleozoic sediments [Alavi, 1996].Volcanism is still sub-active as demonstrated by the quaternary activities around the highest mountain of the Alborz chain, the Damavand volcano. Two different kinds of relative motions control the tectonic activity of this belt [Ritz et al., 2006]: first, a generally compressive motion starting since about 7 Ma (Million years) due to the northward convergence (5 mm/yr) of central Iran toward Eurasia [Vernant et al., 2004]; second, the northwestward motion of the South Caspian Basin with respect to Eurasia resulted in a sinistraltranspressional regime in the Alborz region (4 mm/yr left lateral shear) [Vernant et al., 2004]. However, the tectonic regime in the central Alborz seems to be changing to transtensional due to an acceleration of the northwestward movement of the South Caspian block which started since the middle Pleistocene [Masson et al., 2006;Ritz et al., 2006]. Previous studies [e.g., Asudeh, 1982;Dehghani and Makris, 1984;Tatar, 2001;Ashtari et al., 2005;Doloei and Roberts, 2003;Radjaee et al., 2007;Rastgoo et al., 2018] have described the crustal structure in the Alborz Mountains and reported different values for the Moho depth in Alborz based on different methods. According to these studies, the crustal thickness varies from 34 km in the Tehran region to 55 km beneath central Alborz (See Table 1 for more details).
Along the Alborz mountain belt, seismic activity occurs primarily in the upper crust but with some infrequent events in the lower crust, particularly in the western part of the belt which is called Talesh, where the South Caspian basin underthrusts NW Iran [Engdahl et al., 2006]. Teleseismic waveform modelling of moderate-sized earthquakes reveals centroid depths of up to ∼27 km in the western Alborz and Talesh [Jackson et al., 2002] and a few teleseismic locations

NW Zagros
The Zagros fold belt is the result of the continued convergence between the Arabian and Eurasian plates, which has been in progress since the Miocene episode of continental collision [Stocklin, 1968;Falcon, 1974;Hessami and Jamali, 2006]. The Zagros fold belt with Northwest to Southwest trending of Iran (Almost 1,500 km in length and up to ∼300 km wide) is a major structural element of the Alpine-Himalayan belt. It is one of the most rapidly deforming and seismically active fold-and-thrust belts anywhere in the world [Vernant et al., 2004]. The Zagros region is one of the youngest and most active tectonic regions on Earth so that more than 50% of teleseismic recorded earthquakes in Iran have occurred in this region [Zamani and Agh-Atabai, 2011]. This area is home to some of the world's largest oil reservoirs and was the location of the initial continental collision between the Arabian and Eurasian plates. The area is generally composed of limestone with upper Pliocene units of sandstone [Molnar, 2006].  [Falcon, 1961;Berberian, 1995;Sepehr and Cosgrove, 2004;Emami et al., 2010] (3) the Imbricate Zone (IZ) (also called High Zagros Thrust Belt of Crush Zone), limited by the High Zagros Fault (HZF) [Berberian, 1995] to the southwest, is a highly deformed domain, involving multiple tectonic thrust sheets composed of sedimentary, radiolaritic and ophiolitic rocks, which represents the distal cover rocks of the Arabian Plate, (4) The metamorphic and magmatic Snandaj-Sirjan Zone (SSZ) separated from previous unit by the Main Zagros Fault (MZF). This zone is an Iranian continental block that is thrusted to the SW, on top of the MZF [e.g., Falcon, 1967;Stocklin, 1968], and (5) The Tertiary Urumieh-Dokhtar Magmatic Arc (UDMA) [Varges et al., 2011;Jimenez-Munt et al., 2012;Tunini et al., 2015].
Our study area is located in the northwestern edge of the Simply Folded Belt. The Zagros Simply Folded Belt is located to the south of the High Zagros Fault [Berberian, 1995;Khadivi, 2010). Unlike the other areas of Zagros which have active reverse faults [Berberian, 1995;Jackson and McKenzie, 1984;Nowroozi, 1972;Talebian and Jackson, 2004], the NW Zagros region, has a right lateral strike-slip motion which is confirmed by Fault-plane solutions [Mokhtari et al., 2004].
The strain rate calculated from the historical and recent earthquakes in Zagros is less than 10% of the expected Arabia-Eurasia convergence rate, which shows that 90% of the total deformation energy is being released by aseismic processes [Shoja-Taheri and Niazi, 1981;McKenzie, 1988;Ekstrom and England, 1989]. The responsible for the subdued seismic deformation in Zagros may be the existence of a plastic layer (the Hormoz formation) between the sedimentary column and the basement rock [Jackson and McKenzie, 1988]. Zagros contains a sedimentary cover that spans the entire Phanerozoic and is up to 10-15 km thick [e.g., O'Brien, 1957;James and Wynd, 1965;Stocklin, 1968;Falcon, 1969;Colman-Sadd, 1978]. With a mixture of strong platform carbonates and weaker evaporates, marls and shales, the stratigraphy has long been known to exert an important influence on the style of deformation.
The results show that compared to the Alborz region, the epicentral distribution in Zagros has a weak multifractal (i.e. less heterogeneous) structure. The two distinct multifractal distribution patterns in these regions reflect different underlying seismotectonic processes related to earthquake activity. Both the fact that, in Zagros, there are more small earthquakes than large ones and the relatively low level of discontinuous seismicity with the sporadic occurrence of strong destructive events in Alborz confirm that the Zagros region is relatively more homogeneous than Alborz [Zamani and Agh-Atabai, 2011].
During the last decades, the crustal thickness of the Zagros Mountain belt and its foreland basin has been the topic of numerous studies [Verges et al., 2011;Jimenz-Munt et al., 2012]. The results of several studies [Synder and Barzangi, 1986;Paul et al., 2006Paul et al., , 2010Afsari et al., 2011;Jimenez-Munt et al., 2012;Motaghi, et al., 2014;Tunini, et al., 2015] obtained through several methods show that the crustal thickness varies from 35-45 km in the Mesopotamian Foreland and Arabian Platform to between 44-69 km below the Zagros Mountains with the maximum values beneath the SSZ zone [Tunini et al., 2015]. See Table 1 for more details.
The earthquake of November 12, 2017, with a moment magnitude of 7.3 at an approximately 18 km depth, has been one of the most destructive earthquakes (436 people died according to the latest statistics published) over the past two decades to have happened in NW Zagros [Alavi et al., 2018].

Data
In this study, we use the three components waveforms recorded in 2011-2016 by seismic stations which belongs to the Iranian Seismological Center (IRSC) and the International Institute of Earthquake Engineering and Seismology (IIEES). The IRSC stations consisted of three components short-period and broadband velocity seismometers. The short period seismic stations are equipped with Kenimetrics SS-1 sensors, which their response is flat at the frequency band of 1-25 Hz. In the broad-band stations of IRSC which different seismometers such as Trillium-40s, Trillium-120s, and Güralp-CMG3EPS-120s have been employed, their response is flat at a frequency less than 50 Hz. The broadband IIEES stations are equipped with Güralp CMG-3T seismometers, which their velocity response is flat at the frequency range of 0.0083-50 Hz. The sampling rate of both IRSC and IIEES dataset which was used in this study is 50 samples/s. The details of these stations are provided in Table 2. In the present study, 8717 seismograms of 677 local earthquakes (418 events in Alborz and 259 in NW Zagros) with magnitude M N between 3 and 5.7, were selected. The signal-to-noise ratio of selected seismograms is greater than 3.0. The distribution of seismic stations and events used in this study are shown in Figure 1.
As Figure 1 shows, 23 and 8 seismic stations are located in the Alborz and NW Zagros regions, respectively.
By considering the epicentral distance, we grouped the dataset for R<100 km and 100<R<200 km, which hereafter will be called, dataset-1 and dataset-2, respectively. The magnitude-distance distribution of our dataset is shown in Figure 2. Most of events which were used in this study, have shallow focal depth (1.0 h 20 km), and the average focal depth is about 10 km. Only a few earthquakes have focal depth of more than 20 km, with the deepest earthquake of ~ 28 km depth ( Figure 2).

Q C estimation
In this study, the SBS [Aki and Chouet, 1975] and SIS [Sato, 1977] methods have been used to estimate the coda wave attenuation.

SBS method
The generation and propagation of coda waves may be described by a single backscattering model. This mechanism was proposed by Aki [1969] and Aki and Chouet [1975] to describe the time dependence of the scattered energy density at the source location in the 3-D space [Tuve et al., 2006]. The single backscattering model [Aki and 7 Variation of coda wave attenuation in the Alborz and NW Zagros regions  Table 2. The quality factor at reference frequency 1 Hz, that is, Q 0 , attenuation parameter, n, and the result of statistical analysis including correlation coefficients, R 2 , and standard deviations (presented after ±) for epicentral distance range of 0-100 km and coda window length of 25 s for the SBS and SIS methods, vertical component.   [Rautian and Khalturin, 1978;Roecker et al., 1982;Pulli, 1984] by using local network data. This model is based on the assumption that the source of the earthquake and the receivers are located at the same point in an infinite medium [Ugalde et al., 2002]. According to the SBS model, the coda waves are interpreted as backscattered body waves generated by numerous heterogeneities present in the Earth's crust and upper mantle [Ma'hood, 2014]. The coda amplitudes, A(f, t), in a seismogram can be expressed for a central frequency f over a narrow bandwidth signal, as a function of the lapse time t, measured from the origin time of the seismic event, as: (1) Where f is the frequency, is the geometrical spreading parameter which is considered one of values among 1.0, 0.5 or 0.75 for body waves, surface waves or diffusive waves, respectively [Sato and Fehler, 1998], S(f) is the coda source function at frequency f and is considered a constant, t c is the lapse time and Q c is the quality factor for coda waves. The amplitude of the coda wave, A c (f,t c ), is calculated using the envelope function of the coda amplitude by applying the Hilbert transform according to the following formula: (2) In which x(f,t c ) is the amplitude of the band-pass-filtered coda waves with central frequency f at lapse time t c and H is the Hilbert transform function. The low cut-off and high cut-off for nine passbands with a bandwidth 0.667f are given in Table 3. Coda window seismograms are usually smoothed by applying Root-Mean-Square (RMS) techniques to each envelope [Havskov et al., 1989). Using the coda decay envelope is the most commonly used approach for estimating coda Q [e.g., Woodgold, 1994;Rahimi et al., 2010;Farrokhi et al., 2015;Pezeshk, 2016a, Hovskov et al., 2016].
The equation (1) is valid at lapse times (measured from the earthquake origin time) of approximately t>2t s , in which t s is the travel time of the direct S-wave [Rautian and Khalturin, 1978]. Q c -1 ( f ) can be easily estimated from the recorded seismograms, by fitting the envelopes of the filtered and smoothed seismic traces to the Eq. (1) over a specified time window t 1 t t 2 [Aki and Chouet, 1975;Tuve et al., 2006;Woong Chung et al., 2009]. The start of the time window is taken as t 1 = 2t s and the end of the time window, t 2 , is 2t s +W, which W represents the coda window length.
Variation of coda wave attenuation in the Alborz and NW Zagros regions In general, is assumed to be 1.0 for a short distance within 100 km when the direct Pg and Sg phases arrive, and 0.5 for a distance beyond 100 km when Lg and surface waves are dominant [Chun et al., 1987;Shin and Hermann, 1987;Petukhin et al., 2003]. Most researchers regarded to be a constant beyond 100 km [e.g., Rogers et al., 1987;Brockman and Bollinger, 1992;Kumar et al., 2005;Mukhopadhyay et al., 2006;Rahimi et al., 2010;Farroki et al., 2015;Sedaghati and Pezeshk, 2016a;Dobrynina et al., 2017]. Other researchers treated to be a variable depending on the propagation distance [e.g., Atkinson 2004;Singh et al., 2004;Galluzzo et al., 2015;De Siena et al., 2016;Havskov et al., 2016], in which most of them used the multiple scattering method for estimating of coda Q. In this work, the coda window lengths up to 60 s have been used by applying the single scatting method. Therefore, putting =1 and taking a natural logarithm from both sides of Eq. (1), gives: ( Where b and c are equal to -πf/Q c and ln(S(f )), respectively. The slope of the linear-squares fit between ln[A(f, t)*t] and t yields the Q value for a specific frequency and lapse time window (Figure 3).

SIS Method
The SBS method assumes that the source of the earthquake and the receiver are coincident; therefore, it holds, empirically, for t>2t s , where t s is the S wave travel time [Rautian and Khalturin, 1978]. Sato [1977] developed the SIS model that considers the case of non-coincident source and receiver, thus allowing one to begin the coda analysis just after the S wave arrival. In this model, spherical radiation and isotropic scattering are assumed, and the random distribution of scatters is considered to be homogeneous and isotropic [Ugalde et al., 2002]. The SIS model is particularly important when one is restricted to measurements of coda waves close to the S-wave arrival. For example, in seismically noisy locations the coda wave amplitudes at times greater than twice the S-wave travel time are often less than the background noise level, for small events [Novelo-Casanova and Lee, 1991]. Here, in order to comparison, the same starting point for coda window (2ts) is used in both methods, however, the quality of data (SNR) was good even using t s for starting point in the SIS method. Hence, the natural logarithm of the coda wave envelope is estimated as follows: (4) where A c (f,t c ) represents the observed root-mean-square (RMS) amplitude of the narrow band-pass-filtered waveforms with central frequency f, r is the source-receiver distance, t c is the lapse time measured from the origin The attenuation of coda waves, Q c −1 , can be easily estimated from the slope of the straight line fitting the measured versus t for a given central frequency (Figure 3) [Rautian and Khalturin, 1978].

Lateral variation of Q 0 in Alborz and NW Zagros
The vertical and horizontal Q-factor values have been estimated at each station for 10 frequency bands from the SBS and SIS models so as to investigate the lateral variation of the uppermost part of the lithosphere attenuation structure in our study area which includes two tectonically different regions. Then the frequency-dependent equations of coda quality factor values have been obtained from the power-law equation. The estimated parameters from the power-law equation including the quality factor at reference frequency 1.0 Hz, that is, Q 0 , the frequency dependent parameter, n, and the result of statistical analysis, which include correlation coefficients and standard deviations, are presented in Table 2.
Medium heterogeneity and the small-scale lateral variations in the medium characteristic are expected to be more common in the upper lithosphere [Rahimi et al., 2010;Irandoust et al., 2015]. So, we concentrated on the study of coda waves at shorter coda window lengths and for smaller epicentral distances, because they are less affected by deeper parts of the lithosphere, and the heterogeneity of the upper lithosphere is better manifested in them.
Thus, datasets including the epicentral distance range of 0-100 km are selected for each station and Q c for the coda window length of 25 s was estimated using the SBS and SIS methods. Because of, the observed lateral variation might be due to the difference in the average epicentral distance of the recorded earthquakes in each station.
Therefore, besides Q c at each station, the average epicentral distance has been reported in Table 2. Comparison of the results shows that the obtained Q 0 values using the SIS method are slightly larger than those values obtained by the SBS method (Table 2). The variation of Q 0 values may be due to the differences either in the geological environment or the lithospheric structure in the region. However, in order to discuss the existence of such differences in Alborz, we should be certain that the average epicentral distance of each dataset is the same. So, it is necessary to check the distance parameter because the rays that propagate to longer distances penetrate into greater depths where the attenuation might be different and as a result, two stations with similar attenuation properties could show different Q 0 if one receives rays from a different distance. As shown in Table 2, there is a rough systematic variation between the average distance and Q 0 at all stations except for station GIDE, as the larger Q 0 values belong to stations that recorded more distant events. For example, the station CSN1 with lowest Q 0 has an average distance of 22 km and the station THKV with highest Q 0 in this study has an average distance of 85 km (that is the maximum average epicentral distance among the stations). In station GIDE the coda-Q has estimated using just a single record and so cannot be reliable.
Consequently, it can be said that the observed lateral variation in Alborz is mainly due to the difference in dataset properties. This result is in good agreement with the obtained result for the Central Alborz region by Rahimi et al. [2010]. The calculated n value indicates a very strong frequency dependence of Q c in the Tehran (TEH) station. As mentioned earlier, coda waves are assumed as backscattered body waves generated by the numerous heterogeneities distributed randomly in the Earth's crust and upper mantle. Figure 1 shows that the ray coverage of TEH station is nearly uniform. Therefore, the higher value of n for the coda quality factor may be caused by the direct influence of high heterogeneity level [Langston, 2003] of the Earth's crust and the upper mantle in the Tehran region on coda waves. This area has been confirmed by the dominant existence of folded structures and left-lateral strike-slip and thrust faults parallel to mountain belt trending towards E-W [Jackson et al., 2000;Allen et al., 2003]. Table 2 shows that the average Q 0 value for stations (Except DHR, KER, KCHF, and SNGE) in the Zagros equal or less than the average value for the Alborz region. There is also a rough systematic distance dependency in average Q 0 values for stations of NW Zagros. Hence, we are convinced that the differences in Q 0 values between stations reflect the properties of the dataset, however, the geological effect of the seismic station and ray path is undeniable.

Attenuation variation with depth
The coda quality factor represents the average attenuation property of an ellipsoidal volume with the source and receiver as its focus and depth as its height [Sedaghati and Pezeshk, 2016a]. The size of the study area depends on the epicentral distance and the focal depth of the earthquake, as well as on the coda window length. The seismic wave attenuation and the frequency parameter (n) are also dependent on the coda window length, such that as the coda window length increases, the quality factor increases and the frequency parameter decreases [Aki and Chouet, 1975;Kopnichev, 1991;Rautian andKhalturin, 1978, Sedaghati andPezeshk, 2016a]. Numerous studies in different regions of the world illustrate that as the coda window length increase, so does Q c [Havskov et al., 1989, Gupta et al., 1998, Kumar et al., 2005Rahimi et al., 2010]. As the epicentral distance and the coda window length increase, the deeper areas of the lithosphere play a role in the formation of the coda waves because the waves arriving later in the seismogram can be reflected from the deeper parts of the lithosphere than those arriving earlier.
Therefore, by changing the coda window length, it can be examined the behaviour of Q c (f) as a function of the volume of the presumed region of coda formation and its maximum depth. Thus, that can be interpreted as an increase in the quality factor by increasing depth. Hence, in order to investigate the variation of attenuation with depth, the Q c value was estimated for 11 coda window lengths, which are taken from 10 to 60 s with an increment of 5 s.
In the present study, the obtained results show that the coda seismic quality factor strongly depends on the frequency and coda window length. Despite that the attenuation of seismic waves has been the subject of several studies around the world, its dependence on frequency is still unclear [Dobrynina et al., 2017]. According to Aki and Chouet, [1975], the quality Reza Emami et al.
factor variations with frequency are related to heterogeneities that are randomly distributed in the lithosphere [Dobrynina et al., 2017]. For example, low-frequency signals with longer wavelength are not affected by the small scale heterogeneities [Furumura and Kennett, 2008]. There is a relationship between the frequency dependence of the quality factor and tectonic activity and age of the crust [Mak et al., 2004;Sato and Fehler, 1988].
In addition to the coda window length, the epicentral distance parameter has also been used to check the variations of the coda quality factor with depth, because the stations at shorter epicentral distances sample the coda waves from shorter zones while those at greater epicentral distances cover a larger and deeper area [Rahimi et al., 2010]. Therefore, the data were divided into two datasets based on the epicentral distances R<100 km (dataset-1) and 100 km<R< 200 km (dataset-2). Then, to study the attenuation properties of the study areas with depth in more details the Q c values were estimated from different coda window lengths. The obtained results from various coda windows for the vertical and horizontal components are presented in Tables 4 to 6, respectively. The variation of the average Q 0 (Q c in 1.0 Hz) and n parameters with the coda window length in Alborz and in NW Zagros using the two datasets are shown in Figures 5 and 6. The Q 0 value and the frequency-dependent parameter (n) have an inverse behaviour (Figures 5 and 6).
Additionally, studies show that as the lapse time increases, so does Q 0 [Rocker et al., 1982;Pull, 1984;Woodgold, 1994;Zelt et al., 1999;Sertçelik, 2011]. this difference increases along with the coda window length. The increase rate of Q 0 with the coda window length is not uniform in the second dataset. An interesting point is a great change in the increasing trend of Q 0 in NW Zagros compared to Alborz at 50 s. Thus, the slope of Q 0 for Zagros decreases continually for the coda window length greater than 50 s until the last coda window, 60 s. This decrease is stronger for NW Zagros compared to Alborz for the dataset-2. This means that the lower lithosphere beneath NW Zagros is more attenuative in comparison with the lower lithosphere beneath Alborz, whereas at shallower depths, there is no significant difference between their attenuation characteristics ( Figure 5). As mentioned earlier, the coda waves at a given lapse time are generated by the first order scatters that are located on the surface of an ellipsoid with the earthquake source and the station as foci. The dimension of these ellipsoids depends on the coda window length and epicentral distance. For R<100 km, the dimension of ellipsoids or penetration depths are lower than those for 100 <R<200 km.

Reza Emami et al.
14 At R<100 km, mainly the upper part of the lithosphere, which is no significant difference in parameters effects on Q value in two regions, are scanned. So, a different pattern is not observed on Q 0 versus coda-window-length diagrams at the short distance. But for 100<R<200 Km, the lower part of the lithosphere is scanned which has a significant difference in seismic attenuation characteristics due to the decoupling of Phanerozoic sedimentary phenomena in the basement of Zagros. Therefore a different pattern is observed on Q 0 versus coda-window-length diagrams at the long distance.
Low Q c values characterize high tectonic activity and heterogeneity [Sertçelik, 2011]. Generally, the low value of Q 0 (<200) and high frequency parameter (n) (>0.7) is related to the more seismic activity and heterogeneities in the upper part of the lithosphere [Aki, 2003]. Figures 5 and 6   Note: ΔQ 0 is the standard deviation of Q 0 ; Δn is the standard deviation of the frequency-dependent parameter; A and Z in parentheses show Alborz and NW Zagros regions respectively. W=10s to 164 at W=60 s, and the frequency-dependent (n) varies from 1.04 to 0.72 for the dataset-1 in Alborz region, depending on the coda window length (See Tables 4-6 for full details). Therefore, these results illustrate that the Alborz and NW Zagros regions are tectonically active regions.
An approximate estimation of the maximum depth through the coda window length and the average epicentral distance can be obtained using the Pulli [1984] method based on the single backscattering model proposed by Aki [1969] and Aki and Chouet [1975]. In this model, the estimated attenuation of coda wave is the average decay of amplitudes of backscattered waves on the surface of an ellipsoid volume with the earthquake source and the station as foci [Pulli 1984;Gupta et al., 1998;Rahimi et al., 2010]. In other words, the coda waves at a given lapse time are generated by the first order scatters that are located on the surface of an ellipsoid [Malin, 1978]. The semi-major and semi-minor axes of the surface projected ellipsoid are given by a = V s t/2 and b = (a 2 -(Δ 2 /4)) 1/2 respectively, where V s is S wave velocity, Δ is average epicentral distance and t is the average lapse time [Pulli, 1984]. The maximum depth of volume of ellipsoid from which coda wave generation would occur for different lapse times which is given by h=b+h av , where h av is the average focal depth (~10 km) of events [Pulli, 1984;Havskov et al., 1989;Canas et al., 1995;Rahimi et al., 2010;Sertçelik, 2011]. We have assumed an S-wave velocity of 3.5 km/s based on the previous studies [Rahimi et al., 2010;Motaghi et al., 2014;Irandoust et al., 2015;Rastgoo et al., 2018]. The average lapse time is taken to be t = t start + W/2, where t start is the starting time of the coda window and W is the coda window length. The observed Q c reflects the average attenuation properties of the volume of the ellipsoid.
The analysis of two data sets at different coda window lengths allows us to estimate the coda quality factor at different depths in the study area. As mentioned above, in order to calculate the maximum depth of scatters responsible for the generation of coda waves, for each coda window length and in each dataset it is required to have Δ and t start related to the two datasets including rays with epicentral distances R<100 km and 100<R<200 km in both study areas. The attenuation variation with depth in the two datasets is shown in Figure 7.  Table 7.
The existence of a change in the curve for Alborz in the first dataset can be interpreted as a change in the attenuation structure from depth ~110 km to ~125 km (Figure 7a). This figure shows that the trend of the attenuation curve for the Alborz and NW Zagros regions is similar and decreases by increasing depth, as Q 0 has a steep slope versus depth. This observation implies the existence of a high attenuation (low velocity) anomalous structure beneath the lithosphere of Alborz. Reported lithospheric mantle heterogeneities beneath the Iranian Plateau [Tunini et al., 2015], shows that the Lithosphere-Asthenosphere Boundary (LAB) shallows to ∼120 km beneath Alborz.
The attenuation variation with depth in the second dataset including the epicentral distance 100<R<200 km has been shown in Figure 7b. This dataset shows the average Q 0 over a maximum depth of 145 km. The Q 0 curve in this figure shows that the trend of attenuation in the Alborz and NW Zagros regions is similar and decreases as the Reza Emami et al.
18 depth increases to 180 km. Beneath ~180 km, the trend of Q 0 value for NW Zagros changes and the attenuation increases by increasing depth, so that in this part of the mantle, NW Zagros gets more attenuation, as it decreases less rapidly compared to lithosphere of the Alborz region. This could also indicate the presence of a transparent mantle beneath ~180 km in the study area. The existence of a LAB at 210 km depth beneath Zagros has been reported by Tunini et al. [2015]. The area coverage is computed using the formulation given by Pulli [1984]

Discussion
In this study, coda-Q values have been estimated for the Alborz and NW Zagros regions from the vertical and horizontal components using the SBS [Aki and Choet, 1975] and SIS [Sato, 1977] methods by equations (3) and (4), respectively. The results show that the estimated quality factor using the SIS method is slightly higher than to that value of the SBS method. The only difference between these equations is the Geometrical-Spreading Correction (GSC) factor of coda wave amplitudes, which are t and r/√k(a) for the SBS and SIS methods, respectively. Therefore, the estimated Q c value only depends on the calculated GSC values by the SBS and SIS methods. The coda-Q values are proportional to the slope of geometrical correction. When comparing the normalized amplitude (NA) of geometrical correction it was shown that the NA values of the SIS method have slightly higher slopes compared with the SBS method, which gives higher values for estimated Q c at short lapse times, t c, less than 85 s [Farrokhi et al., 2015].
Due to the scattering nature of the coda waves, they come from all directions [Hovskov et al., 2016). Therefore, we would not expect any difference between the components [Del Pezzo et al., 1985;Sato et al., 2012]; hence, all the three components can be used. Furthermore, in order to have a large number of observations, all three components should be used [Hovskov et al., 2016]. The estimated results using the vertical and horizontal components are tabulated in Tables 4 to 6.
In this study, the relationship between the obtained quality factors for the two regions and the abovementioned properties were investigated.
As discussed earlier, for R<100 km the coda waves attenuation in Zagros is slightly higher than in Alborz, but it isn't significant and are pretty similar ( Figure 5). Our results are in good agreement with those obtained in previous studies [Ghasemi et al., 2009;Sedaghati and Pezeshk, 2016b;Darzi et al., 2019]  (ANOVA) technique. They concluded that there is no significant difference between the patterns of the attenuation PGAs (Peak Grand Accelerations) and SAs (Spectral Acceleration) with distance for the Zagros and Alborz regions.
The coda lapse time is related to the region of coda wave sampling. Longer lapse times represent longer travelled distances and larger depth intervals [Sertçelik, 2011;Irandoust et al., 2015]. Recent studies [e.g., Shapiro et al., 2000] have revealed that at short distances the scattering attenuation is dominant while at larger distances the intrinsic absorption which is very sensitive to the temperature is dominant [Gao, 1992;Havskov et al., 2016]. Seismic attenuation is a complicated process, which can be caused or affected by earth properties such as seismicity, temperature, the thickness of sediments, rock types, and the age of the earth. It is noteworthy that, the strong effect of temperature on seismic attenuation has been calculated in a number of experimental studies, and it was used recently to correlate seismic attenuation anomalies in the mantle with temperature variations [e.g., Durak et al., 1993;Romanowicz, 1994;Mitchell, 1995;Romanowicz, 1995;Artemieva et al., 2004].
Due to the similarity of the average temperature of the two regions up to a depth of approximately 100 km (Tunini et al., 2015), the slightly higher attenuation at shallower depths in the Zagros region can be attributed to the extremely high seismicity and large sedimentary cover beneath Zagros compared to Alborz. Zagros shows the thickest sedimentary basin of Iran (Teknik and Ghods, 2017). Therefore, the stratigraphic thickness reaches to a maximum of 14 km in the NW of Zagros [Molinaro et al., 2005;Casciello et al., 2009;Nissan et al., 2011]. While the stratigraphic thickness of sediments deposited from the Precambrian to the early Cenozoic reaches to a maximum of 10 km in Alborz [Assereto, 1966]. The volcano-clastic sedimentary column is exposed mostly in the southern hills of the Alborz Mountains with a maximum stratigraphic thickness of 6 km [Teknik and Ghods, 2017].
If we want to compare the seismicity of the two regions, we can say that Zagros is more active. The Zagros belt is a weak lithosphere, incapable of sustaining high strain levels and a very heterogeneous stress system. Most earthquakes occur in such weak crusts, and their occurrence has often been attributed to the movements of small fault segments (Zamani and Agh-Atabai, 2011). It is the seismically most active region of Iran, such that more than 50% of the teleseismic recorded earthquakes in Iran have occurred in the Zagros region [Mirzaei et al., 1998;Zamani and Agh-Atabai, 2011]. Like Zagros though to a smaller degree, Alborz is a zone of high seismicity, and strong earthquakes have been recorded in the area. The strongest earthquake recorded is the Damghan earthquake of M S 7.9, which occurred on 22 Dec 856 and caused 200,000 fatalities [Ambraseys and Melville, 1982].
As Figure 5 shows, unlike R<100 km, the difference of the obtained coda waves attenuation for both study areas increases and becomes much stronger while the length of coda window increases for the second dataset. This suggests that the lithosphere beneath NW Zagros is more attenuative compared to that below Alborz. The significant difference between Q c values in the two regions for 100<R<200 is interesting but this is explainable. According to the suggested model of Mohammadi et al., [2013], the remnant of the fossil Neo-Tethyan subduction exists beneath Zagros and SSZ in depth range of 100-150 km. The combination of our results with Mohammadi et al., [2013] emphasizes low velocity zone exists may be due to the partial melting of the remnant of the fossil Neo-Tethyan. The average temperature of the two regions are similar from the surface to a depth of 100 km [Tunini et al., 2015] but for much of the Phanerozoic, Zagros was a subsiding continental margin which was accumulating a thick package of sediments. Basins that were filled with sediments of low conductivity are able to retain high temperatures long after initial lithospheric stretching increases the geothermal gradient [McKenzie, 1981;Jackson, 1987], suggesting that the Arabian basement underlying the sediments of Zagros may remain hot and weak [Nissen et al., 2011]. The increased temperature of the basement of Zagros as a result of the decoupling of Phanerozoic sedimentary maybe have an important role in decreasing the quality factor of seismic waves.

Comparison of Q c with other observations
In this study, Coda-Q values and its frequency-dependent relation were estimated for different coda window lengths using the equation Q=Q 0 f n for Alborz and NW Zagros. The results showed that Alborz and NW Zagros regions are in tectonically active regions. In the last decade, some research has been undertaken to estimate the Q c frequency relations at different tectonic regions of Iran [e.g., Rahimi et al., 2010;Rahimi and Hamzehloo, 2008;Ma'hood and Hamzehloo, 2009;Farrokhi et al., 2015;Irandoust et al., 2015 andRahimi et al., 2009]. Rahimi et al. [2010] have evaluated a relation Q c =79f 1.07 with a coda window length of 25 s, in the central Alborz region which is the central part of our study area. The dataset used in Rahimi et al. [2010] consisted of 345 local earthquakes (2<M<5) with the epicentral distance less than 100 km. We derived Q c =109f 0.87 at same coda window length in the Alborz region.
for Rahimi and Hamzehloo (2008), whereas epicentral distances used in this study range from 4.5 to 100 km.
According to the Q c frequency relation provided by Irandoust et al. [2015], Q 0 and n are 109 and 0.99, respectively, in SE Zagros (the Bandar-Abbas region) at the coda window length of 30 s. The epicentral distance of events (2.5<M<5) used in their study are less than 100 km. In the present study, the obtained relation for Q c for NW of Zagros is Q c = 108f 0.86 . It is obvious that the Q 0 and n values given by them are similar and greater, respectively, to those results of our estimates. As mentioned earlier, the parameter n is indicative of the degree of heterogeneity of the crust [Aki, 1981], therefore, it can be said that NW Zagros is more homogeneous in compared to SE Zagros.
Irandoust et al. [2015] also estimated Q c = 83f 1.08 for north of Zagros, that have an overlap with our study area.
The difference in the obtained results of the two studies can be due to the distinct part of the studies areas as well as different datasets. Table 8 (upper part) presents Q 0 and n values and Figure 8 shows the comparison coda Q variation with those of other studied in different regions of Iranian Plateau.
We have compared some of the results of Q c studies obtained by various researchers in the world [Sertçelik, 2012;Mukhopadhyay et al., 2006;Mak et al., 2004;Birch et al., 2015;Sedaghati and Pezeshk, 2016a;Singh and Hermann, 1983]. The obtained values are comparable to the Q c frequency trend of the Northwestern Himalayas (Q c =113f 1.01 ); the East Anatolia Fault Zone, Turkey (Q c = 58f 0.82 ); the New Madrid Seismic Zone (Q c =598 f 054 ), Hong Kong (Q c = 256f 0.7 ); Kaapvaal craton, South Africa (Q c =327f 0.81 ), Northeast United States (Q c = 900f 0.35 ), and Central United States (Q c = 1000f 0.20 ). Table 8 (lower part) compares the Q-factor function obtained in this study compared to those from the other regions in the world. As shown in Figure 9, the Alborz and NW Zagros regions obviously follow the trend for regions with active seismic activities.

Conclusion
The quality factor (Q c ) and the frequency-dependent parameter (n), in the lithosphere of the Alborz and NW Zagros regions, were estimated from an analysis of the coda waves of 677 local events which were recorded by short period and broadband seismic stations of IRSC and IIEES. In this study, the vertical and horizontal components of seismograms have been analysed to study the lateral and depth variation of coda waves attenuation using the SBS and SIS methods. In conclusion, the following results have been reached: In this study, Q c =109±7f 0.87±0.04 and Q c = 108±6f 0.86±0.03 were estimated using the vertical component seismograms of dataset-1 at a coda window length of 25 s, for the Alborz and NW Zagros regions, respectively. The average Q c for the Alborz and NW Zagros regions were evaluated Q c =148±13f 0.81±0.05 and Q c = 124±5f 0.89±0.03 , respectively, by using dataset-2.These results derived for both dataset by applying the SBS method.
By applying the SIS model for the dataset-1, the obtained Q c in the Alborz and NW Zagros regions are Q c =111±8f 0.87±0.04 and Q c = 111±6f 0.86±0.03 respectively. Also, by using the dataset-2, the average Q c =149±13f 0.81±0.05 and Q c = 126±5f 0.89±0.03 were estimated for the Alborz and NW Zagros regions, respectively.
The estimated Q 0 <200 and n>0.7 through different coda window lengths in the study regions illustrate that the Earth's crust and upper mantle beneath the Alborz and NW Zagros are seismotectonically active and heterogeneous.
Increasing Q 0 and decreasing n parameter with coda window lengths or penetration-depths, shows that the heterogeneity level decreases with depth in both Alborz and NW Zagros regions. Rapid increasing of estimated Q 0 values for Alborz in comparison with those in Zagros, suggests that the lower lithosphere beneath NW Zagros is more attenuative.
As mentioned previously, the seismic wave attenuation was obtained from all three components. Estimated Q c values using different components show that there is no significant difference for using vertical and horizontal seismograms, so, it can be concluded that the degree of heterogeneity is same in the vertical and horizontal directions are essentially identical.