# Frequency-dependent Lg attenuation in Northeast China and its implications

Frequency-dependent Lg attenuation in Northeast China and its implications Summary Lg attenuation and site responses in Northeast (NE) China are estimated by implementing the Reverse Two Station Method (RTSM) at narrow band central frequencies of 0.5, 1.0, 2.0 and 3.0 Hz using 453 earthquakes recorded by 201 seismic stations deployed in the region from 1995 to 2013. The RTSM has the advantage of removing source and site effects without requiring a priori models. Tomographic images produced at a resolution of 2° × 2° at all frequencies exhibit a high degree of lateral variation of Lg attenuation in NE China. The Great Xing’an, Lesser Xing’an and Songen-Zhangguangcai Ranges show Q values above 400 at all frequencies. At central frequencies of 0.5 and 1 Hz, Sanjiang Basin Songliao Basin, Erlian Basin and Hailar Basin consistently show Q values lower than 400. Holocene and Pleistocene volcanoes, including the Wudalianchi and Jingpuhi volcanic fields, also appear as regions of low Q (<400) at 0.5 and 1 Hz and at higher frequencies the effects of volcanoes diminish. At high frequencies (≥2 Hz), the sedimentary basins show Q values higher than 400 and overall Q values increase with frequency in NE China, thus obeying a power-law frequency dependence. A linear regression of frequencies in the range of 0.25–4.5 Hz results in the parameters describing the power-law frequency dependence of the region, with an average Q0 value of 428 and a frequency-dependent factor (η), describing the strength of dependence, of 0.68. Overall, Lg attenuation in NE China appears to be due to thick late Cretaceous sediments, Holocene and Pleistocene volcanism, moderate to high heat flow, partial melts and variation in the thickness of the crustal wave guide. The site responses calculated at 0.5, 1.0, 2.0 and 3 Hz show a high degree of lateral variation as well as variation with frequency in NE China. At 0.5 and 1.0 Hz, the Great Xing’an and Lesser Xing’an Ranges show deamplification while the Hailar, Erlian and southern Songliao basins and the Songen-Zhangguangcai Range show amplification. At higher frequencies (≥2 Hz), the pattern nearly reverses, with deamplification east of the Songliao Basin and amplification to the west of the Basin. The deamplification observed east of the Songliao Basin could be caused by basaltic lava flows resulting from Cenozoic volcanism in the region. Asia, Guided Waves, Seismic attenuation, Seismic tomography, Site effects 1 INTRODUCTION The tomographic mapping of crustal attenuation is an important method to understand regional crustal structure. Crustal attenuation is especially sensitive to sediments, fluids, partial melt and sudden variation in crustal thickness. Furthermore, its calibration is important for nuclear monitoring purposes and magnitude determination. Here, we examine the attenuation of the Lg wave in Northeast (NE) China using data from the NorthEast China Extended SeiSmic (NECESS) array experiment and other seismic experiments conducted in the region from 1995 to 2014. Lg is the most prominent high frequency wave that can be observed at regional distances and can be modelled as a superposition of multiple S wave reverberations and SV to P and P to SV conversions within the crustal waveguide (Bouchon 1982; Herrmann & Kijko 1983; Olsen et al. 1983; Campillo et al.1984; Kennett 1986; Ou & Herrmann 1990) or as the sum of higher mode surface waves (Oliver et al. 1955; Knopoff et al. 1973; Kennett 1984; Bostock & Kennett 1990). The average group velocity window of Lg ranges from 3.0 to 3.6 km s−1 and most of its energy lies within the 0.2 to 10 Hz frequency band in NE China. The efficient propagation of the Lg wave is highly dependent on the thickness of the crustal waveguide. In a simplified layer over a half space crustal model, the maximum number of viable modes necessary to generate Lg is proportional to the depth of the wave guide (Aki & Richards 1980; Zhang & Lay 1995). High frequency (>1 Hz) Lg ground motion at long distances (>200 km) is caused by supercritically reflected S waves bouncing back and forth between the crustal waveguide (e.g. Bouchon 1982; Herrmann 1985; Kennett 1986) as well as from supercritical reflections at internal interfaces (Burger et al. 1987). Using synthetic seismograms generated from ray theory, main Lg ground motion at regional distances has been shown to be caused by supercritically reflected S waves confined within the waveguide between the free surface and crustal interfaces below the source (Ou & Herrmann 1990). Thus, the Lg wave propagates efficiently over continental paths, having been observed at distances up to 30° in the Canadian Shield (Hasegawa 1985), which have sufficient crustal thickness. In contrast, very thin oceanic crust (<10 km) is unable to maintain these modes (e.g. Press & Ewing 1952; Knopoff et al. 1979; Kennett 1986; Zhang & Lay 1995), resulting in the complete blockage of Lg in oceanic crust (Cao & Muirhead 1993; Zhang & Lay 1995). This dependence of the modes on crustal thickness also results in the Lg wave quickly losing energy while propagating through areas that undergo rapid changes in crustal thickness, such as the transition that occurs beneath Tibet (e.g. McNamara et al. 1996; Xie 2002; Fan & Lay 2003). Lg attenuation has been shown to be sensitive to crustal attenuation structure (e.g. Baqer & Mitchell 1998), intrinsic properties of the crustal wave guide such as crustal temperature (e.g. Frankel 1991), high heat flow (e.g. Aleqabi & Wysession 2006), fluid content (e.g. Mitchell 1995), the presence of partial melts (e.g. Xie et al. 2004), unconsolidated sediments (e.g. Mitchell & Hwang 1987), crustal deformation (e.g. Pasyanos et al. 2009) and the age of the last crustal deformation event (e.g. Mitchell et al. 1997). NE China is a region of high geopolitical and tectonic interest. Only a few studies of regional seismic wave attenuation have been conducted in this region (Rapine & Ni 2003; Zhao et al. 2010) prior to the deployment of the NECESS array in 2009. Zhao et al. (2010) applied a single-station, single-event method to simultaneously invert for frequency-dependent Lg Q and the source function in NE China using 20 stations and 125 events. They found the area has an average Q0 (Q at 1 Hz) value of 414 and observed that the sedimentary basins have lower Q values while granitic mountain ranges were found to have higher Q values. Ranasinghe et al. (2015) used 78 crustal earthquakes recorded by the 127 station NECESS array to study Lg attenuation at 1 Hz in the region using the Reverse Two Station Method (RTSM). They observed that regions with little to no sediment (<1 km) tended to be characterized by high Q values (>800) while regions with moderately high heat flow (>70 mW m−2) tended to be characterized by low Q values (<800). Furthermore, they concluded Holocene volcanoes have a greater effect on Lg attenuation than Pleistocene volcanoes in the region. In an update of our prior study, we discuss results obtained from the RTSM for NE China, including estimates of frequency-dependent Q, and site responses. In this study, we address anomalies observed in the Lg attenuation tomography of NE China and their relationship to geology, sediment thickness, crustal thickness, and heat flow using a larger data set than our prior study and we also study frequency-dependent Lg attenuation and site responses. Comparisons with previous studies in the region will also be made. Finally, we will discuss the frequency dependence of Lg attenuation and the site responses observed in NE China. Our results in NE China confirms the frequency dependence of Lg attenuation that has been observed in other parts of the world (e.g. Chun et al.1987; Erickson et al.2004), which could be due to the effects of transitioning from a scattering-dominated regime to an intrinsic-dominated attenuation regime (Dainty 1981). 2 GEOLOGY AND TECTONICS OF NE CHINA Present day NE China is a geologically and structurally complex region composed of deep, organic-rich sedimentary basins, mountain ranges, and Pleistocene and Holocene volcanoes (Fig. 1). The Songliao, Hailar, Erlian, Sanjiang and Bohai Basins represent the major sedimentary basins in the region. The Great Xing’an, Lesser Xing’an and Songen-Zhangguangcai Ranges are located between the basins and they are mostly composed of granitic and volcanic rocks. Changbaishan, a Holocene volcanic complex located on the border between China and North Korea, is formed from a subduction-induced upwelling (Tang et al. 2014). Other Holocene volcanic fields include the Wudalianchi, Jingpuhi and Longgan volcanic complexes (Fig. 1). Figure 1. View largeDownload slide A geological map of northeast China, dashed black line represents the boundaries of the tomographic images. Black lines are active faults and dashed black lines are Late Triassic sutures. Red triangles represent volcanoes of Holocene age and open triangles represent volcanoes of Pleistocene age. Abbreviations on the maps are: BHB, Bohai Basin; CBV, Changbaishan Volcanic Field; DMF, Dunhua-Mishan Fault; EB, Erlian Basin; GB, Gobi Basin; GXR, Great Xing’an Range; HB, Hailar Basin; JMS, Jiayin-Mudanjiang Suture; JPV, Jingpuhi Volcanic Field; JYF, Jiamusi-Yitong Fault; LGV, Longgan Group of Volcanoes; LXR, Lesser Xing’an Range; NF, Nenjiang Fault; SLB, Songliao Basin; SNB, Sanjiang Basin; SS, Suolon Suture; SZR, Songen-Zhangguangcai Range; UBB, Upper Bureya Basin; WVF, Wudalianchi Volcanic Field; YTB, Yitong Basin; ZBB, Zeya–Bureya Basin. Figure 1. View largeDownload slide A geological map of northeast China, dashed black line represents the boundaries of the tomographic images. Black lines are active faults and dashed black lines are Late Triassic sutures. Red triangles represent volcanoes of Holocene age and open triangles represent volcanoes of Pleistocene age. Abbreviations on the maps are: BHB, Bohai Basin; CBV, Changbaishan Volcanic Field; DMF, Dunhua-Mishan Fault; EB, Erlian Basin; GB, Gobi Basin; GXR, Great Xing’an Range; HB, Hailar Basin; JMS, Jiayin-Mudanjiang Suture; JPV, Jingpuhi Volcanic Field; JYF, Jiamusi-Yitong Fault; LGV, Longgan Group of Volcanoes; LXR, Lesser Xing’an Range; NF, Nenjiang Fault; SLB, Songliao Basin; SNB, Sanjiang Basin; SS, Suolon Suture; SZR, Songen-Zhangguangcai Range; UBB, Upper Bureya Basin; WVF, Wudalianchi Volcanic Field; YTB, Yitong Basin; ZBB, Zeya–Bureya Basin. The Songliao Basin is composed of a series of interconnected half-grabens characterized by a thin crust (25–35 km) and relatively high heat flow (70 mWm−2). The basement of the Songliao Basin is cut by a system of Jurassic rifts, which are filled with volcanic rocks interlayered with Late Jurassic coal-bearing continental clastics (Zhou et al. 1985; Ulmishek 1992). In the Middle Cretaceous, during the early sag development stages of the Songliao Basin, sediment deposition was mainly confined to rift structures (Ulmishek 1992). In the Late Cretaceous, sediments were deposited in a large area extending up to 200 000 km2 from the centre. During this time, the deposition environment changed from an anoxic deepwater environment to a shallow oxic environment. Thus, shales dominate the central part of the basin while deltaic sandstones dominate the northern, southern and western margins (Ulmishek 1992). During the Cenozoic, NE China was characterized by extensive alkaline volcanism, the subduction of the Pacific plate underneath NE China, and the opening of the Sea of Japan. Volcanism in NE China was initially confined to the interior of the Songliao Basin and later (∼ 28 Ma) migrated to the flanks (Liu et al. 2001). Volcanic activity peaked in the Neogene along the Tan-Lu fault (TLF) and the Jiamusi-Yitong Fault (JYF; Liu et al. 2001) east of the Songliao Basin. Present day volcanism is dominated by the Changbaishan, Jingpuhi, and Wudalianchi volcanic complexes. Northeast China is currently undergoing east-west compression from the end of the opening of the Sea of Japan and the forces exerted by the subduction of the Pacific plate (Sagiya et al. 2000; Jin et al. 2007). 3 DATA PROCESSING The seismic waveform data used for this study were collected from 201 seismic stations operated in NE China and surrounding regions from January 1995 to December 2013 (Fig. 2). We used stations from several networks (Supporting Information Table S1), including: (1) 127 temporary broad-band stations from the NECESS Array (2011–2013), (2) 15 broad-band stations from the Incorporated Research Institute for Seismology (IRIS) Global Seismic Network (GSN) (1995–2013), (3) 19 broad-band stations from the NE China seismic experiment (1998–1999), (4) 26 broad-band stations from the natural and man-induced seismicity in the Yanquing-Hualia Basin and the Heicheng Area, China experiment (2002–2008), (5) 12 broad-band stations from the China National Seismic Network (2009–2013), (6) 1 broad-band station from the Korean Seismic Network (2013–2013), and (7) 1 broad-band station from the Japan meteorological agency seismic network (2007–2013). Figure 2. View largeDownload slide A map of the study region showing topography and the locations of all the crustal earthquakes used to study Lg attenuation. Magenta, black, red and purple circles represent the 453 earthquakes used to study Lg attenuation. The blue triangles represent the locations of the Northeast China Extended Seismic (NECESS) Array stations. The inverted magenta triangles represent the location of the Global Seismic Network (GSN) stations used in this study. The red stars represent the location of China National Seismic network stations. The purple diamonds represent the location of the NE China Seismic Experiment stations. The cyan stars represent the locations of the Natural & Man-Induced Seismicity in the Yanquing-Hualia Basin and the Heicheng Area, China experiment seismic stations. Finally, the red and cyan diamonds represent Korean Seismic Network and Japanese Metrological Agency seismic stations, respectively. Figure 2. View largeDownload slide A map of the study region showing topography and the locations of all the crustal earthquakes used to study Lg attenuation. Magenta, black, red and purple circles represent the 453 earthquakes used to study Lg attenuation. The blue triangles represent the locations of the Northeast China Extended Seismic (NECESS) Array stations. The inverted magenta triangles represent the location of the Global Seismic Network (GSN) stations used in this study. The red stars represent the location of China National Seismic network stations. The purple diamonds represent the location of the NE China Seismic Experiment stations. The cyan stars represent the locations of the Natural & Man-Induced Seismicity in the Yanquing-Hualia Basin and the Heicheng Area, China experiment seismic stations. Finally, the red and cyan diamonds represent Korean Seismic Network and Japanese Metrological Agency seismic stations, respectively. The sampling rate of the waveform data gathered by stations other than the NECESS Array network were adjusted to match the 40 samples s−1 rate of the NECESS Array stations prior to further processing. A list of the stations used in this study and their initial sampling frequency is included in the supporting material of this paper. The preliminary earthquakes chosen for this study range in magnitude from 2.5 to 7.0 (magnitude types are given in Supporting Information Table S2) and event depths were restricted to a maximum depth of 40 km, while epicentral distances were limited to a minimum of 2.5° and a maximum of 20°. A choice of focal depth of 40 km would make earthquakes more likely to be crustal in nature. The epicentral distance limits are in place because Lg does not develop prior to a distance of at least 2.5° while beyond 20° the Lg phase can no longer be distinguished from the pre-event noise in NE China. The waveform data were first processed by temporarily filtering the vertical component seismograms with a two-pass, three-pole recursive Butterworth filter with corners at 0.2 and 5 Hz for visualization purposes. The Pn arrival time was then picked to define the 20 s long pre-Pn noise window, with the end of the window starting at the pick. The Lg group velocity window was initially set to range from 3.6 km s−1 to 3.0 km s−1, as a visual guide to pick the onset of the Lg arrival. Then, the beginning of the Lg group velocity window was visually picked on each seismogram. If the characteristic Lg waveforms were absent, the seismograms were removed from further processing. After visually picking Pn and Lg arrival times, we chose 453 earthquakes (Fig. 2) with at least 3 clear Lg waveforms for an event (Supporting Information Table S2). The individual spectra were calculated by applying a Fourier transform with a 20 per cent cosine taper to the chosen pre-Pn noise window and Lg windows to avoid spectral leakage (Xie 1998). The signal-to-noise ratio (SNR) for Lg was calculated by dividing the respective Root Mean Square (RMS) of the Lg spectra by the RMS of the pre-Pn noise spectra in the frequency domain. Any records with an average SNR below 2.0 were removed from further processing. Finally, the instrument responses were removed from both the noise and signal spectra using instrument response information obtained from the IRIS Data Management Center (DMC). 4 METHOD 4.1 Amplitude parametrization The spectral amplitude of a seismic wave can be defined as   $$A\left( {f,d} \right) = S\left( f \right)I\!\left( f \right)E\!\left( f \right)G\!\left( d \right){\rm{exp}}\left( {\frac{{ - \pi fd}}{{vQ\left( f \right)}}} \right),$$ (1)where A(f, d) is the amplitude at a spectral frequency of f measured a given distance d from the source, S( f ) is the source response, I( f ) is the instrument response, E( f ) is the site response, G(d) is the geometrical spreading, $${\rm{exp}}\, ( {\frac{{ - \pi fd}}{{vQ( f )}}} )$$ is the frequency-dependent attenuation term, Q( f ) is the frequency-dependent quality factor, and v is the assumed average group velocity of 3.5 km s−1 for Lg. The geometric term for Lg is assumed to be in the form of   $$G\left( d \right) = {G_0}\ {d^{ - m}}$$ (2)where G0 is a constant (Chun et al.1987; Yang 2002; Bao et al.2011) and m is the geometric spreading coefficient. In this study, we implement the RTSM (Chun et al. 1987) to solve for the frequency-dependent quality factor Q(f), which describes Lg attenuation in NE China. Rather than invert for the geometric spreading coefficient m (Ranasinghe et al.2015), we use the generally accepted value of 0.5 (e.g. Yang 2002) in order to be consistent with previous Lg attenuation tomographic studies in the region (Zhao et al. 2010) and elsewhere (e.g. Xie 2002). 4.2 RTSM In the RTSM (Figs 3a and b), spectral amplitudes recorded at two stations (i and j) from two events (a and b) that lie along a great circle arc are divided out to remove the source response, site response and G0:   \begin{eqnarray} &&{\frac{{{A_{ai}}\left( {f,d} \right)}}{{{A_{aj}}\left( {f,d} \right)}}\ \frac{{{A_{bj}}\left( {f,d} \right)}}{{{A_{bi}}\left( {f,d} \right)}} = {\left( {\frac{{{d_{ai}}\ {d_{bj}}}}{{{d_{aj}}{d_{bi}}}}} \right)^{ - m}} } \nonumber\\ &&{\quad \times \, {\rm{exp}}\left[ {\frac{{\pi f}}{{vQ\left( f \right)}}\left( {{d_{aj}} - {d_{ai}} - {d_{bj}} + {d_{bi}}} \right)} \right]} \end{eqnarray} (3)where Aai and Aaj are the spectral amplitudes recorded by stations i and j respectively for the event a. Similarly, Abi and Abj are spectral amplitudes recorded by stations i and j for the event b. dai, daj, dbi and dbj are the corresponding epicentral distances as shown in Figs 3(a) and (b). In the RTSM we allow for a maximum variation of 15° from the great circle arc between the stations and the events, as shown in Fig. 3(b) (Der et al. 1984). The back azimuthal angle is also restricted to a maximum of 15° in order avoid possible obliqueness in longer ray paths. The path Q values are inverted following Xie & Mitchell (1990) to obtain the interstation Q values. Lg attenuation has been shown to follow a power law (e.g. Chun et al.1987; Erickson et al.2004) and it is defined as   $$Q\left( f \right) = {Q_0}{\left( {\frac{f}{{{f_0}}}} \right)^\eta }$$ (4)where, Q0 is the reference Q value at 1 Hz,  f0 is a reference frequency of 1 Hz and η is the frequency-dependent factor. The η represent variation in Q value with the frequency. Frequency-dependent pathQ values for the RTSM can be given as   $$\ln \left( {\frac{1}{{Q\left( f \right)}}} \right) = \ln \left( {\frac{1}{{{Q_0}}}} \right) + \ \eta {\rm{\ ln}}\left( {\frac{{{f_0}}}{f}} \right).$$ (5) Figure 3. View largeDownload slide An illustration of the Reverse Two Station Method (RTSM) configuration. The red stars a and b represent earthquakes and the black triangles i and j represent seismic stations. (a) The ideal geometry for the RTSM configuration. (b) The actual geometry for the RTSM configuration. The distance between stations i and j is represented by dij. The epicentral distances between the earthquakes a and b and the stations i and j are represented as dai, daj, dbi and dbj, respectively. θa and θb are the subtended angles between earthquakes a and b for stations i and j, respectively. Figure 3. View largeDownload slide An illustration of the Reverse Two Station Method (RTSM) configuration. The red stars a and b represent earthquakes and the black triangles i and j represent seismic stations. (a) The ideal geometry for the RTSM configuration. (b) The actual geometry for the RTSM configuration. The distance between stations i and j is represented by dij. The epicentral distances between the earthquakes a and b and the stations i and j are represented as dai, daj, dbi and dbj, respectively. θa and θb are the subtended angles between earthquakes a and b for stations i and j, respectively. The η and Q0 values for the RTSM are found using the intercept and the slope of the linear regression of eq. (5). 4.3 Site response The RTSM can also be used to calculate the relative site responses of individual stations (e.g. Bao et al. 2012; Gallegos et al.2017). These site responses are found by multiplying the amplitude ratios rather than dividing them as in eq. (3), resulting in eq. (6) below   \begin{eqnarray} \ln \ \left( {\frac{{{E_{i}}\left( f \right)}}{{{E_j}\left( f \right)}}} \right) &=& \frac{{{d_{aj}} - {d_{ai}}}}{{{d_{aj}} + {d_{bi}} - {d_{ai}} - {d_{bj}}}}\ \ln \left( {\frac{{{A_{ai}}\left( {f,d} \right)d_{ai}^m}}{{{A_{aj}}\left( {f,d} \right)d_{aj}^m}}} \right) \nonumber\\ &&+ \ \frac{{{d_{bi}} - {d_{bj}}}}{{{d_{aj}} + {d_{bi}} - {d_{ai}} - {d_{bj}}}}\ln \left( {\frac{{{A_{bi}}\left( {f,d} \right)d_{bi}^m}}{{{A_{bj}}\left( {f,d} \right)d_{bj}^m}}} \right). \nonumber\\ \end{eqnarray} (6)where Ei and Ej are the site responses of station i and  j, respectively. Eq. (6) can be formulated as a least squares problem. As the first step in calculating the site responses, eq. (6) is inverted using the LSQR algorithm (Paige & Saunders 1982). The site responses are calculated relative to a set of reference stations. These reference stations were varied to test the robustness of the model. Stations that had sampled a large number of ray paths had robust solutions that did not vary with the choice of reference stations. Thus, we simply chose the reference stations with the highest number of ray paths (NE21, NE23, NE62, NE76, NE77, NE82 and NE91). 5 RESULTS AND DISCUSSION 5.1 Lg tomography Utilizing RTSM, Lg Q tomography images (Fig. 4) with 2° × 2° resolution (Fig. 5) were produced at frequency bands of 0.25–0.75 Hz, 0.5–1.5 Hz, 1.0–3.0 Hz and 1.5–4.5 Hz with central frequencies of 0.5, 1.0, 2.0 and 3.0 Hz, respectively. The tomography and resolution maps were restricted to only include the area ranging from 42.5° to 48° latitude and from 116° to 135° longitude to minimize the edge effects from areas with low (<10) ray-path density (Fig. 6). We also did not include Lg path Q values less than 15 and greater than 2000 in the inversion, considering them outliers. A Q error ratio is also calculated by dividing the path Q value by its path Q error, whereas path Q error is one standard deviation of the best fit between the path Q values and frequency in each frequency spectra. Path Q error ratios greater than one were also removed from the inversion. We used a damping value of 0.75 and a Gaussian smoothing to keep the inversion stable and avoid having the edge effects bleed into the tomography images. Figure 4. View largeDownload slide Tomographic maps of Lg Q in NE China created using the RTSM at central frequencies of: (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz. The Q value increases with increasing frequency. Figure 4. View largeDownload slide Tomographic maps of Lg Q in NE China created using the RTSM at central frequencies of: (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz. The Q value increases with increasing frequency. Figure 5. View largeDownload slide 2° × 2° RTSM Lg Q resolution tests using a damping value of 0.15 and 15 per cent normally distributed noise at central frequencies of (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz. Lg attenuation is well-resolved at all frequencies with the exception of smearing at the edges. The Changbaishan and Wudalianchi volcanic fields are poorly resolved. Figure 5. View largeDownload slide 2° × 2° RTSM Lg Q resolution tests using a damping value of 0.15 and 15 per cent normally distributed noise at central frequencies of (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz. Lg attenuation is well-resolved at all frequencies with the exception of smearing at the edges. The Changbaishan and Wudalianchi volcanic fields are poorly resolved. Figure 6. View largeDownload slide RTSM Lg ray-path density at central frequencies of (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz using 3018, 4539, 4032 and 2724 ray paths, respectively. The best ray-path coverage is observed in the central part of NE China. A lack of ray-path coverage can be observed at the edges, particularly in the Changbaishan region and south of the Suolon Suture. Figure 6. View largeDownload slide RTSM Lg ray-path density at central frequencies of (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz using 3018, 4539, 4032 and 2724 ray paths, respectively. The best ray-path coverage is observed in the central part of NE China. A lack of ray-path coverage can be observed at the edges, particularly in the Changbaishan region and south of the Suolon Suture. The tomography images show increasing average Q values of 281, 411, 684 and 931 at 0.5, 1.0, 2.0 and 3.0 Hz, respectively. Uncertainties of the models were calculated by inverting path Q error values in the same fashion as the path Q values (Fig. 7). The best ray-path density is obtained in the Songliao Basin and the Great Xing’an Range (Fig. 6). Although a larger dataset was used in this study compared to Ranasinghe et al. (2015), it did not result in a substantial increase in the area of the ray-path coverage (Fig. 6) because of the restrictive RTSM geometry. However, the larger dataset did result in an overall increase in the number of ray paths used in this study (Supporting Information Table S3). The Songliao, Sanjiang, Hailar and Erlian Basins show low (< 400) Lg Q values at 0.5 Hz. The central part of the Songliao Basin, which is dominated by thick shale deposits, exhibits lower Q values compared to the edges of the basin, which are dominated by thin sand stone deposits (Ulmishek 1992). In a laboratory experiment, Winkler and Nur (1979) demonstrated that Q values increase as pore fluid is flushed out of a sand stone by increasing pressure. Therefore, in more compact sediments, the Q value is expected to be higher than in uncompacted, fluid-filled sediments (Mitchell & Hwang 1987). At 0.5 Hz, Sanjiang, Hailar and Erlian basins also show low Q values due to unconsolidated sediments. High attenuation experienced in sedimentary basins could also be due to fluids (Mitchell and Hwang 1987) and focusing and defocusing effects of the basins. The North and central Great Xing’an Ranges show high (>800) Q values and the southern Great Xing’an Range shows moderately high (400 to 800) Q values. The Lesser Xing’an Range and Songen-Zhangguangcai Ranges show moderately high Q values. Overall at 0.5 Hz, sedimentary basins with unconsolidated sediments and the Wudalianchi Volcanic Field show high attenuation and granitic mountain ranges show low attenuation. The results are similar at 1 Hz frequency, but there is reduced attenuation beneath the Greater Xing’an Range and along the Nenjiang Fault and increased attenuation in the east beneath the Lesser Xing’an Range and the southeastern Jiamusi-Yitong and Dunhua-Mishan Faults. Previous Lg attenuation studies have also indicated regions with active faulting, fluid filled fractures and suture zones tend to show both inefficient Lg propagation as well as high Lg attenuation (e.g. Crampin 1994; Mitchell 1995). Sandvol et al. (2001) observed inefficient Lg propagation in the Bitlis Suture and the Zagros fold and thrust belt in the Middle East. In another study in Tibet, Bao et al. (2012) found anisotropic 1/QLg is strongest in the northwestern Sonpan-Ganzi fold belt with high Q values running parallel to major strike-slip fault planes in the region. At 1 Hz, the Songen-Zhangguangcai and Great Xing’an Ranges show high Q values of 800–1000. Altogether, at 1 Hz sedimentary basins and the Wudalianchi and Jingpuhi Volcanic Fields show high attenuation while the Songen-Zhangguangcai and Great Xing’an Ranges show low attenuation. Figure 7. View largeDownload slide Lg Q errors calculated at (a) 0.5 Hz, (b) 1.0 Hz, (c) 2.0 Hz and (d) 3.0 Hz frequencies show most of the errors are less than 0.4 in areas with a high ray-path coverage. Errors are large at the edges of the tomographic images. Figure 7. View largeDownload slide Lg Q errors calculated at (a) 0.5 Hz, (b) 1.0 Hz, (c) 2.0 Hz and (d) 3.0 Hz frequencies show most of the errors are less than 0.4 in areas with a high ray-path coverage. Errors are large at the edges of the tomographic images. At higher frequencies, all the mountain ranges show high Q values and the Q values in the sedimentary basins increase compared to lower frequencies (≤1 Hz). The low Q values observed at low frequencies (≤1 Hz) correspond with thick unconsolidated sediments (Fig. 8), volcanic fields and fault lines found throughout NE China. At 2 Hz, the average Q has increased but Q variations are subdued apart from a high attenuation patch in the southeast corner and by 3 Hz the image is dominated by high Q values in the region. Figure 8. View largeDownload slide Lg site response maps of NE China overlaid on a sediment thickness (Laske & Masters 1997) map at central frequencies of (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz. The sizes of the triangles and the circles represent the values of the deamplification and the amplification, respectively. At central frequencies of 0.5 Hz and 1 Hz, most of the seismic stations in the basins show amplification. Additionally, at central frequencies of 2 Hz and 3 Hz, most of the seismic stations east of the Songliao except in the Sanjiang Basin show deamplification. Figure 8. View largeDownload slide Lg site response maps of NE China overlaid on a sediment thickness (Laske & Masters 1997) map at central frequencies of (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz. The sizes of the triangles and the circles represent the values of the deamplification and the amplification, respectively. At central frequencies of 0.5 Hz and 1 Hz, most of the seismic stations in the basins show amplification. Additionally, at central frequencies of 2 Hz and 3 Hz, most of the seismic stations east of the Songliao except in the Sanjiang Basin show deamplification. Both Pleistocene and Holocene volcanoes seem to have a large effect on Lg attenuation at lower (≤1 Hz) frequencies and at higher frequencies (≥2 Hz) both Pleistocene and Holocene volcanoes exhibit high Q values. Similar observations of Lg attenuation have been made in other parts of the world where Neogene and Cenozoic volcanism cause poor Lg propagation in middle east (e.g. Sandvol et al.2001) and high attenuation in the western U.S. (e.g. Phillips & Stead 2008; Gallegos et al. 2017). High Q values exhibited by the Quaternary volcanoes at higher frequencies could be due to less scattering experienced at higher frequencies. The Lg Q tomography maps created at different frequencies are consistent with previous models of Lg attenuation in NE China (Zhao et al. 2010) as well as with the geology in the region. These maps show an increase in Q values with frequency, that is, a power-law relationship (eq. 4) as predicted in other studies (e.g. Chun et al.1987; Erickson et al.2004; Zhao et al. 2010). We obtained a frequency-dependent factor (η value) of 0.68 and an average Q0 (Q at 1 Hz) value of 428 in NE China (Fig. 9). Figure 9. View largeDownload slide Plot of $$\ln ( {\frac{1}{Q}} )$$versus $$\ln ( {\frac{{{f_0}}}{f}} )$$ using the RTSM. Q0 and η values calculated from the linear fit for the RTSM are 0.68 and 428, respectively. Figure 9. View largeDownload slide Plot of $$\ln ( {\frac{1}{Q}} )$$versus $$\ln ( {\frac{{{f_0}}}{f}} )$$ using the RTSM. Q0 and η values calculated from the linear fit for the RTSM are 0.68 and 428, respectively. Comparing our Lg attenuation results with the geology of the region, we find that at low frequencies (≤1 Hz), sedimentary basins exhibit high attenuation (Q < 400) while mountain ranges show low attenuation (Q > 400). In the case of the basins, high attenuation is partially due to the presence of porous, fluid rich sediments and focusing/defocusing effects. These results are consistent with the previous Lg attenuation results at a central frequency of 1 Hz reported by Ranasinghe et al. (2015), where regions with thick sediments (>1 km) were characterized by low Q0 values (< 800). We also compared our Lg attenuation results to a sediment thickness map (Fig. 8) created using the global model made by Laske & Masters (1997). The map shows that the thickest sediments (∼7 km) are found in the central part of the Songliao Basin. The attenuation images produced at 0.5 Hz and 1.0 Hz show low attenuation (Q < 400) in areas with thick (>1 km) sediments. The attenuation images produced at 1.0 Hz show high attenuation in the Songliao, Hailar, Erlian and Sanjiang Basins. At central frequencies of 2 and 3 Hz, the thick sediments appear to cause very little attenuation, which could also be related to the increase of Q values in the basins with frequency. Overall, sediments are responsible for some of the high attenuation observed in the basins at lower frequencies. High heat flow has been also identified as a major cause of Lg attenuation in various parts of the world, such as the Tibetan Plateau (Xie 2002) and the Basin and Range province (Aleqabi & Wysession 2006). We compare our results with a heat flow study conducted by Wang and Cheng (2012) using a limited number of heat flow data values in the region. The authors found medium to high heat flow within the Songliao Basin, which corresponds to (70 mW m−2) an estimated average temperature of 600° C at the Moho beneath the basin. The high Lg attenuation found in the Songliao Basin at lower frequencies (≤1 Hz) could be partially due to the high heat flow, but there is insufficient heat flow data to conclusively state high heat flow is the main cause for the high attenuation observed in the Songliao Basin. Strong Lg attenuation and blockage has been observed in continental areas that undergo rapid changes in crustal thickness over relatively small distances, such as the Tibetan Plateau (Bàth 1954; Ni & Barazangi 1983; McNamara et al. 1996; Fan & Lay 2003), in the Bolivian Altiplano (Baumont et al. 1999) and in the European Alpine mountains (Campillo et al. 1993). We also compared our attenuation results with a crustal thickness map of NE China (Tao et al. 2014) created using the receiver functions (Supporting Information Fig. S1). The crustal thickness underneath Sanjiang Basin, Jingpuhi and Wudalianchi Volcanic Fields and the majority of Songliao Basin, with the exception of the northern portion, varies from ∼20 to 32 km. The crustal thickness in rest of NE China varies from ∼ 32 up to 45 km. We observed regions with thin crust, especially at the central frequency of 1 Hz, were found to show high attenuation but overall we did not observe a one to one relationship between crustal thickness and Lg attenuation. Upper crustal shear (S) wave maps produced at 1, 4, 6 and 7 km depths in NE China by Li et al. (2016) using joint inversion of Rayleigh wave ellipticity and phase velocity from the ambient noise and earthquake data collected from the NECESS Array show a broad low velocity zone in the northern part of the Songliao Basin and another low velocity zone near southwestern edge of the Basin as well as in Erlian, Hailar and Sanjiang Basins (Supporting Information Fig. S2). The low attenuation regions in the mountainous regions of the 1 Hz Lg attenuation image also seems to correspond with high velocity zones in the upper crustal shear wave maps created at 1, 4, 6, and 7 km depth. The low velocity zones at 1 and 4 km depth also corresponds with high Lg attenuations regions in our 0.5 and 1 Hz Lg attenuation images. Low S-wave velocities tend to correspond with low Lg Q values in the upper crust and it could be due to unconsolidated sediments causing low Q as well as low S-wave velocities in the upper crust. Zhao et al. (2010) used 1720 Lg spectra derived from 20 broad-band stations and 125 events to create Lg attenuation images in NE China at central frequencies of 0.5, 1.0 and 2 Hz using a single-event, single-station method (Supporting Information Fig. S3). They observed low Q values in the Bohai, Songliao, Hailar, Erlian, Sanjiang and Zeya-Buriya Basins and high Q values in the Great Xing’an, Lesser Xing’an and Songen-Zhangguangcai Ranges. Our RTSM results agree well with the broad regional features of Zhao et al. (2010). Being able to produce quite similar Lg attenuation tomography models implementing two different techniques and utilizing two different data sets can be taken as a strong measure of the robustness of both models. However, the Lg Q values calculated by Zhao et al. (2010) are frequently lower by ∼200–300 than those we calculated in our study. The difference between the Q values may have arisen from the fact that in our study we inverted for Lg Q−1, while Zhao et al. (2010) inverted for Q. By inverting for Lg Q−1 we avoid the potential bias introduced by averaging Lg Q values, which can be negatively affected by the presence of very large Lg Q values. Zhao et al. (2010) also observe an increase of Lg Q values with frequency, which is consistent with our study. Comparing with our previous study of Lg attenuation in NE China at a central frequency of 1 Hz (Ranasinghe et al.2015), our new models created at 0.5, 1.0, 2.0 and 3.0 Hz provide a broader view of Lg attenuation in NE China at different central frequencies. The 1 Hz images of the two studies show features such as sedimentary basins and Holocene volcanoes as high attenuation regions and granitic mountain ranges as low attenuation regions. The resulting Q values of the two studies cannot be compared directly, as the geometric spreading values used in the two studies are different from each other. 5.2 Frequency dependence of Lg In order to model the frequency dependence of Lg attenuation in NE China, the frequency dependence factor (η) and the Q0 values for the RTSM (eq. 5) were calculated plotting $${\log _e}(\frac{1}{Q})$$ against $${\log _e}( {\frac{{{f_0}}}{f}} )$$ using the average Q values found from the tomographic images produced at central frequencies of 0.5, 1.0, 1.5, 2.0, 2.5 and 3.0 Hz corresponding to narrow band frequencies of 0.25–0.75, 0.5–1.5, 0.75–2.25, 1.5–3.0, 1.75–3.75 and 2.0–4.5, respectively (Fig. 9). We obtained an η value of 0.68 and a Q0 of 428 for Lg attenuation in NE China using RTSM. Our results are also consistent with previous Lg attenuation studies by showing an increase in average Q value in the frequency range from 0.5 to 3.0 Hz (e.g.; Erickson et al.2004; Zhao et al. 2010). Frequency-dependent Lg attenuation can be interpreted based on the definition of apparent Q as a linear combination of scattering Q and intrinsic Q (Dainty 1981). Dainty (1981) put forth the idea that Lg attenuation is dominated by scattering Q at lower frequencies (≤1 Hz) and by intrinsic Q at higher frequencies (>1 Hz) by intrinsic Q. The intrinsic Q value stays roughly constant throughout the frequency range while the importance of scattering decreases with an increase in frequency. Under this assumption, the observed power law behaviour of frequency is a result of the contributions from both intrinsic Q and scattering Q. The attenuation maps created at 0.5 and 1 Hz are dominated by scattering attenuation and the maps created at 2 and 3 Hz are dominated by intrinsic attenuation. It should be noted that the true mechanisms behind scattering and intrinsic attenuation are still under debate. The explanation put forth by Dainty (1981) is just one possibility. 5.3 Site response for Lg Frequency-dependent Lg site responses were calculated using eq. (6) at central frequencies of 0.5, 1.0, 2.0 and 3.0 Hz (Fig. 8). At 0.5 Hz for Lg, seismic stations in the Great Xing’an and Lesser Xing’an Ranges consistently show deamplification (negative site response relative to mean) while stations in the Hailar and Sanjiang Basins as well as the southern Songliao Basin and Songen-Zhangguangcai Range show amplification (positive site response relative to mean). However, this pattern nearly reverses at higher frequencies, with deamplification to the east of the Songliao basin and amplification to the west now being observed. At 3 Hz the site responses have the largest effect. Many of the stations located over areas with the thinnest sediments (∼0.5–1.0 km) show the largest changes in amplification with frequency. The station log-gains have values from −0.5 to +0.5. For an average Q of 400 over a 200 km ray-path segment, we expect close to this value for the log-amplitude ratio. At central frequencies of 2 and 3 Hz, stations in the mountain ranges east of the Songliao Basin show consistent and dramatic deamplification. Data shows that signal exists up to 5 Hz and these stations include both permanent and temporary stations. Thus, the deamplification cannot be attributed to data quality or deployment issues and we conclude the effect is due to local site geology. The region does have numerous Cenozoic volcanoes, granitic rocks and basalt flows, and it also lies between Dunhua-Mishan Fault and Jiamusi-Yitong Fault. Phillips and Aki (1986) observed the strongest Lg coda site deamplification for the frequency range of 1.5–3.0 Hz in the granitic Gabilan Range and meta-sediment Franciscan basement sites in California. They also observe site amplification is inversely proportional to deposit age at sedimentary sites. They further observed moderate site amplification in Miocene and older Cenozoic sediment deposits. Thus, we conclude the strong deamplification observed east of the Songliao Basin at higher frequencies is partly due to Cenozoic volcanism. However, we also note that deamplification at these higher frequencies cannot be solely attributed to sedimentation or surface rock type. Lg site responses in the Erlian Basin, which is composed of half graben systems (Changsong et al. 2001), vary drastically between frequencies while most of the stations outside the Erlian Basin show high amplification at 2 Hz. It could be related to near surface features and focusing-defocusing effects from grabens and half-grabens present in the Basin. Akyol et al. (2013) also observed ground amplification in a horst-graben system in western Turkey due to focusing-defocusing effects. They also observed large seismic amplitude variations in seismic stations closer to the basin edges due to the complex nature of the wave propagation. Theoretically, it has also been shown that elastic focusing/defocusing effects in a 3-D Earth model with lateral variations in velocity and anelastic attenuation causes perturbations in the amplitude of surface waves (e.g. Zhou 2009; Ruan and Zhou 2012). Therefore, much of the amplification observed at the edges of the basin could be attributed to focusing and defocusing effects. 6 CONCLUSION The RTSM was used to measure lateral variations in Lg Q in NE China using an extensive data set collected from 1995 to 2013 in the region by various seismic networks. This method eliminates most of the effects of both the source and site terms. Lg Q maps were created applying the RTSM method at central frequencies of 0.5, 1.0, 2.0 and 3.0 Hz which can resolve 2° × 2° features, with the best resolution found in the Great Xing’an Range and Songliao Basin. Average Lg Q values obtained for NE China increase with frequency and we obtained a frequency-dependent factor (η) of 0.68 and an average Q0 (Q value at 1 Hz) of 428 for RTSM. A high degree of lateral variation in Lg attenuation was found in the region at all frequencies. At low frequencies (≤1 Hz), the south and eastern part of the Songliao Basin, as well as the Erlian, Hailar and Sanjiang Basins show high attenuation. The Great Xing’an Range, Lesser Xing’an Range and Songen-Zhangguangcai Range show low attenuation at low frequencies. At low frequencies, areas of high Lg attenuation were found to correspond to thick sediments and at higher frequencies ( ≥ 2 Hz) the correlation ceased to exist. Also at low frequencies, high attenuation corresponds with low shear velocity, high heat flow and Quaternary volcanism in the area. Regions with low attenuation correspond to mountain ranges and areas that have not experienced recent tectonic activity. The high attenuation observed at lower frequencies is more dominated by scattering processes rather than intrinsic attenuation. Lg site responses are found to vary laterally as well as with frequency throughout NE China. At all frequencies, the majority of seismic stations lying in the mountain ranges show deamplification. Amplifications in the basins vary widely with frequency. The observed seismic amplification in the basins could be related to focusing-defocusing effects. The deamplification observed east of the Songliao Basin at higher frequencies (≥2 Hz) could be related to Cenozoic volcanism in the region. ACKNOWLEDGEMENTS We would like to thank the NECESS Array (doi:10.7914/SN/YP_2009) team from China, Japan and the USA for collecting most of the data used in this study. We also used seismic data collected by the China National Seismic Network (doi:10.7914/SN/CB), the Global Seismograph Network (doi:10.7914/SN/IU, doi:10.7914/SN/II), the New China Digital Seismograph Network (doi:10.7914/SN/IC), the NE China Seismic Experiment, the Natural and Man-Induced Seismicity experiment in the Yanquing-Hualia Basin and the Heicheng Area, China (doi:10.7914/SN/XG_2002), and finally the Korean Seismic Network and Japan Metrological Agency seismic network. We would also like to thank the Incorporated Research Institutions for Seismology (IRIS) Data management System (DMC) for allowing us to access the data and promptly answering the queries regarding the data. Comments and suggestions from Jack Xie and another anonymous reviewer have been a great help in improving the manuscript. Finally, we would like to thank Paul Wessel and Walter H.F. Smith for the use of the Generic Mapping Tools (GMT) software, Lawrence Livermore for the use of their Seismic Analysis Code (SAC), the National Geophysical Data Center of NOAA for making the Etopo1 global relief map freely available and Fugro Robertson, Ltd for making the basin classification database freely available through AAPG. New Mexico State University is supported under National Science Foundation (NSF) EAR-0608629, U.S. Air Force Research Laboratory contract No. FA9453-12-C-0235 and Los Alamos National Security Subcontract No. 218475. REFERENCES Aki K., Richards P.G. 1980. Quantitative Seismology , 2nd edn, pp. 251– 253, W.H. Freeman and Co. Akyol N., Kurtulmuş T.Ö., Çamyildiz M., Güngör T., 2013. Spectral ratio estimates for site effects on the horst-graben system in West Turkey, Pure appl. Geophys. , 170( 12), 2107– 2125. https://doi.org/10.1007/s00024-013-0661-2 Google Scholar CrossRef Search ADS   Aleqabi G.I., Wysession M.E., 2006. QLg distribution in the basin and range province of the Western United States, Bull. seism. Soc. Am. , 96( 1), 348– 354. https://doi.org/10.1785/0120040086 Google Scholar CrossRef Search ADS   Bao X., Sandvol E., Ni J., Hearn T., Chen Y.J., Shen Y., 2011. High resolution regional seismic attenuation tomography in eastern Tibetan Plateau and adjacent regions, Geophys. Res. Lett. , 38( 16), L16304, doi:10.1029/2011GL048012. https://doi.org/10.1029/2011GL048012 Google Scholar CrossRef Search ADS   Bao X., Sandvol E., Chen Y.J., Ni J., Hearn T., Shen Y., 2012. Azimuthal anisotropy of Lg attenuation in eastern Tibetan Plateau. J. geophys. Res. , 117( B10), B10309, doi:10.1029/2012JB009255. https://doi.org/10.1029/2012JB009255 Google Scholar CrossRef Search ADS   Baqer S., Mitchell B., 1998. Regional variation of Lg coda Q in the continental United States and its relation to crustal structure and evolution, Pure appl. Geophys. , 153( 4), 613– 638. https://doi.org/10.1007/s000240050210 Google Scholar CrossRef Search ADS   Bàth M., 1954. The elastic waves Lg and Rg along Euro-Asiatic paths, Ark. Geofys. , 2, 295– 324. Baumont D., Paul A., Beck S., Zandt G., 1999. Strong crustal heterogeneityin the Bolivian Altiplano as suggested by attenuation of Lg waves, J. Geophys. Res. , 104( B9), 287– 305. Google Scholar CrossRef Search ADS   Bostock M.G., Kennett B.L.N., 1990. The effect of 3-D structure on Lg propagation patterns, Geophys. J. Int. , 101( 2), 355– 364. https://doi.org/10.1111/j.1365-246X.1990.tb06574.x Google Scholar CrossRef Search ADS   Bouchon M., 1982. The complete synthesis of seismic crustal phases at regional distances, J. geophys. Res. , 87( B3), 1735– 1741. https://doi.org/10.1029/JB087iB03p01735 Google Scholar CrossRef Search ADS   Burger R.W., Somerville P.G., Barker J.S., Herrmann R.B., Helmberger D.V., 1987. The effect of crustal structure on strong ground motion attenuation relations in eastern North America, Bull. seism. Soc. Am. , 77, 420– 439. Campillo M., Bouchon M., Massinon B., 1984. Theoretical study of the excitation, spectral characteristics and geometrical attenuation of regional seismic phases, Bull. seism. Soc. Am. , 74, 79– 90. Campillo M., Feignier B., Bouchon B., Be'thoux N., 1993. Attenuation of crustal waves across the Alpine range, J. geophys. Res. , 98( B2), 1987– 1996. https://doi.org/10.1029/92JB02357 Google Scholar CrossRef Search ADS   Cao S., Muirhead K.J., 1993. Finite difference modelling of Lg blockage, Geophys. J. Int. , 115( 1), 85– 96. https://doi.org/10.1111/j.1365-246X.1993.tb05590.x Google Scholar CrossRef Search ADS   Changsong L., Eriksson K., Sitian L., Yongxian W., Jianye R., Yanmei Z., 2001. Sequence architecture, depositional systems, and controls on development of lacustrine basin fills in part of the Erlian Basin, northeast China, Am. Assoc. Pet. Geol. Bull. , 85, 2017– 2043. Chun K.-Y., West G.F., Kokoski R.J., Samson C., 1987. A novel technique for measuring Lg attenuation results from eastern Canada between 1 to 10 Hz, Bull. seism. Soc. Am. , 77, 398– 419. Crampin S., 1994. The fracture criticality of crustal rocks, Geophys. J. Int. , 118( 2), 428– 438. https://doi.org/10.1111/j.1365-246X.1994.tb03974.x Google Scholar CrossRef Search ADS   Dainty A.M., 1981. A scattering model to explain seismic Q observations in the lithosphere between 1 and 30 Hz, Geophys. Res. Lett. , 8( 11), 1126– 1128. https://doi.org/10.1029/GL008i011p01126 Google Scholar CrossRef Search ADS   Der Z.A., Marshall M.E., O’Donnell A., McElfresh T.W., 1984. Spatial coherence structure and attenuation of the Lg phase, site effects, and interpretation of the Lg coda, Bull. seism. Soc. Am. , 74, 1125– 1148. Erickson D., McNamara D., Benz H., 2004. Frequency-dependent Lg Q within the continental United States, Bull. seism. Soc. Am. , 94( 5), 1630– 1643. https://doi.org/10.1785/012003218 Google Scholar CrossRef Search ADS   Fan G.W., Lay T., 2003. Strong Lg attenuation in the Tibetan Plateau, Bull. seism. Soc. Am. , 93( 5), 2264– 2272. https://doi.org/10.1785/0120030052 Google Scholar CrossRef Search ADS   Frankel A., 1991. Mechanisms of seismic attenuation in the crust: scattering and anelasticity in New York State, South Africa, and southern California, J. Geophys. Res. , 96, 6269– 6289. Google Scholar CrossRef Search ADS   Gallegos A., Ranasinghe N., Ni J., Sandvol E., 2017. Lg attenuation, frequency dependence and relative site response of the western United States as revealed by the EarthScope Transportable Array, Geophys. J. Int. , 209( 3), 1955– 1971. https://doi.org/10.1093/gji/ggx145 Google Scholar CrossRef Search ADS   Hasegawa H.S., 1985. Attenuation of Lg waves in the Canadian Shield, Bull. seism. Soc. Am. , 75, 1569– 1582. Herrmann R.B., 1985. An extension of random vibration theory estimates of strong ground motion to large distances, Bull. seism. Soc. Am. , 75, 1447– 1453. Herrmann R.B., Kijko A., 1983. Modeling some empirical vertical component Lg relations, Bull. seism. Soc. Am. , 73, 157– 171. Jin S., Park P.-H., Zhu W., 2007. Micro-plate tectonics and kinematics in Northeast Asia inferred from a dense set of GPS observations, Earth planet. Sci. Lett. , 257( 3–4), 486– 496. https://doi.org/10.1016/j.epsl.2007.03.011 Google Scholar CrossRef Search ADS   Kennett B.L.N., 1984. Guided wave propagation in laterally varying media—I. Theoretical development, Geophys. J. Int. , 79( 1), 235– 255. https://doi.org/10.1111/j.1365-246X.1984.tb02853.x Google Scholar CrossRef Search ADS   Kennett B.L.N., 1986. Lg Waves and structural boundaries, Bull. seism. Soc. Am. , 76, 1133– 1141. Knopoff L., Schwab F., Kausel E., 1973. Interpretation of Lg, Geophys. J. Int. , 33( 4), 389– 404. https://doi.org/10.1111/j.1365-246X.1973.tb02375.x Google Scholar CrossRef Search ADS   Knopoff L., Mitchel R.G., Kauself E.G., 1979. A search for the oceanic Lg Phase, Geophys. J. Int. , 56( 1), 211– 218. https://doi.org/10.1111/j.1365-246X.1979.tb04776.x Google Scholar CrossRef Search ADS   Laske G., Masters G., 1997. A global digital map of sediment thickness, EOS, Trans. Am. geophys. Un. , 78, F483. Li G., Chen H., Niu F., Guo Z., Yang Y., Xie J., 2016. Measurement of Rayleigh wave ellipticity and its application to the joint inversion of high-resolution S wave velocity structure beneath northeast China, J. geophys. Res. , 121( 2), 864– 880. https://doi.org/10.1002/2015JB012459 Google Scholar CrossRef Search ADS   Liu J., Han J., Fyfe W.S., 2001. Cenozoic episodic volcanism and continental rifting in northeast China and possible link to Japan Sea development as revealed from K-Ar geochronology, Tectonophysics , 339( 3–4), 385– 401. https://doi.org/10.1016/S0040-1951(01)00132-9 Google Scholar CrossRef Search ADS   McNamara D.E., Owens T.J., Walter W.R., 1996. Propagation characteristics of Lg across the Tibetan Plateau, Bull. seism. Soc. Am. , 86( 2), 457– 469. Mitchell B.J., 1995. Anelastic structure and evolution of the continental crust and upper mantle from seismic surface wave attenuation, Rev. Geophys. , 33( 4), 441– 462. https://doi.org/10.1029/95RG02074 Google Scholar CrossRef Search ADS   Mitchell B.J., Hwang H., 1987. Effect of low Q sediments and crustal Q on Lg Attenuation in the United States, Bull. seism. Soc. Am. , 77( 4), 1197– 1210. Mitchell B.J., Pan Y.P., Xie J., Cong L., 1997. Lg coda Q variation across Eurasia and its relation to crustal evolution, J. Geophys. Res. , 102( B10), 22 767– 22 779. Google Scholar CrossRef Search ADS   Ni J., Barazangi M., 1983. High-frequency seismic wave propagation beneath the Indian Shield, Himalayan Arc, Tibetan Plateau and surrounding regions: high uppermost mantle velocities and efficient Sn propagation beneath Tibet, Geophys. J. Int. , 72( 3), 665– 689. https://doi.org/10.1111/j.1365-246X.1983.tb02826.x Google Scholar CrossRef Search ADS   Oliver J., Ewing M., Press F., 1955. Crustal structure of the Arctic regions from the Lg Phase, Bull. geol. Soc. Am. , 66( 9), 1063– 1074. https://doi.org/10.1130/0016-7606(1955)66%5b1063:CSOTAR%5d2.0.CO;2 Google Scholar CrossRef Search ADS   Olsen K.H., Braile L.W., Stewart J.N., 1983. Modeling short-period crustal phases (p, Lg) for long-range refraction profiles, Phys. Earth planet. Inter. , 31( 4), 334– 347. https://doi.org/10.1016/0031-9201(83)90093-6 Google Scholar CrossRef Search ADS   Ou G.B., Herrmann R.B., 1990. A statistical model for ground motion produced by earthquakes at local and regional distances, Bull. seism. Soc. Am. , 80( 6A), 1397– 1417. Paige C.C., Saunders M.A., 1982. LSQR: An algorithm for sparse linear equations and sparse least squares, ACM Trans. Math. Softw. , 8( 1), 43– 71. https://doi.org/10.1145/355984.355989 Google Scholar CrossRef Search ADS   Pasyanos M.E., Matzel E.M., Walter W.R., Rodgers A.J., 2009. Broad-band Lg attenuation modelling in the Middle East, Geophys. J. Int. , 177( 3), 1166– 1176. https://doi.org/10.1111/j.1365-246X.2009.04128.x Google Scholar CrossRef Search ADS   Phillips W.S., Aki K., 1986. Site amplification of coda waves from local earthquakes in central California, Bull. seism. Soc. Am. , 76, 627– 648. Phillips W.S., Stead R.J., 2008. Attenuation of Lg in the western US using the USArray, Geophys. Res. Lett. , 35( 7), L07307, doi:10.1029/2007GL032926. https://doi.org/10.1029/2007GL032926 Google Scholar CrossRef Search ADS   Press F., Ewing M., 1952. Two slow surface waves across North America, Bull. seism. Soc. Am. , 42, 219– 228. Ranasinghe N.R. et al.  , 2015. Lg attenuation in northeast China using NECESSArray data, Geophys. J. Int. , 200( 1), 67– 76. https://doi.org/10.1093/gji/ggu375 Google Scholar CrossRef Search ADS   Rapine R.R., Ni J.F., 2003. Propagation Characteristics of Sn and Lg in Northeastern China and Mongolia, Bull. seism. Soc. Am. , 93( 2), 939– 945. https://doi.org/10.1785/0120010157 Google Scholar CrossRef Search ADS   Ruan Y., Zhou Y., 2012. The effects of 3-D anelasticity (Q) structure on surface wave amplitudes, Geophys. J. Int. , 189( 2), 967– 983. https://doi.org/10.1111/j.1365-246X.2011.05356.x Google Scholar CrossRef Search ADS   Sagiya T., Miyazaki S., Tada T., 2000. Continuous GPS array and present-day crustal deformation of Japan, Pure appl. Geophys. , 157, 2303– 2322. Sandvol E. et al.  , 2001. Tomographic imaging of Lg and Sn propagation in the Middle East, Pure appl. Geophys. , 158( 7), 1121– 1163. https://doi.org/10.1007/PL00001218 Google Scholar CrossRef Search ADS   Tang Y. et al.  , 2014. Changbaishan volcanism in northeast China linked to subduction-induced mantle upwelling, Nat. Geosci. , 7( 6), 470– 475. https://doi.org/10.1038/ngeo2166 Google Scholar CrossRef Search ADS   Tao K. et al.  , 2014. Crustal structure beneath NE China imaged by NECESSArray receiver function data, Earth planet. Sci. Lett. , 398, 48– 57. https://doi.org/10.1016/j.epsl.2014.04.043 Google Scholar CrossRef Search ADS   Ulmishek G., 1992. Geology and hydrocarbon resources of onshore basins in eastern China, U.S. Geological Survey open-file report; 93–4, pp. 48– 57, Denver, CO, USA. Wang Y., Cheng S.-H., 2012. Lithospheric thermal structure and rheology of the eastern China, J. Asian Earth Sci. , 47, 51– 63. https://doi.org/10.1016/j.jseaes.2011.11.022 Google Scholar CrossRef Search ADS   Winkler K., Nur A., 1979. Pore fluids and seismic attenuation in rocks, Geophys. Res. Lett. , 6( 1), 1– 4. https://doi.org/10.1029/GL006i001p00001 Google Scholar CrossRef Search ADS   Xie J., 1998. Spectral inversion of Lg from earthquakes: a modified method with application to the 1995, western Texas earthquake sequence, Bull. seism. Soc. Am. , 88, 1525– 1537. Xie J.K., 2002. Lg Q in the eastern Tibetan Plateau, Bull. seism. Soc. Am. , 92( 2), 871– 876. https://doi.org/10.1785/0120010154 Google Scholar CrossRef Search ADS   Xie J.K., Mitchell B.J., 1990. A back-projection method for imaging large-scale lateral variations of Lg coda Q with application to continental Africa, J. geophys. Res. , 100, 161– 181. Xie J., Gok R., Ni J., Aoki Y., 2004. Lateral variations of crustal seismic attenuation along the INDEPTH profiles in Tibet from Lg Q inversion, J. Geophys. Res. , 109, B10308, doi:10.1029/2004JB002988. Yang X., 2002. A numerical investigation of Lg geometrical spreading, Bull. seism. Soc. Am. , 92( 8), 3067– 3079. https://doi.org/10.1785/0120020046 Google Scholar CrossRef Search ADS   Zhang T.-R., Lay T., 1995. Why the Lg phase does not traverse oceanic crust, Bull. seism. Soc. Am. , 85( 6), 1665– 1678. Zhao L.-F., Xie X.-B., Wang W.-M., Zhang J.-H., Yao Z.-X., 2010. Seismic Lg-wave Q tomography in and around Northeast China, J. geophys. Res. , 115, B08307, doi:10.1029/2009JB007157. https://doi.org/10.1029/2009JB007157 Zhou Y., 2009. Surface-wave sensitivity to 3-D anelasticity, Geophys. J. Int. , 178, 1403– 1410. https://doi.org/10.1111/j.1365-246X.2009.04230.x Google Scholar CrossRef Search ADS   Zhou Z., Songyu Q., Changnian H., 1985. Songliao Basin, in ESCAP Atlas of Stratigraphy IV, Peoples's Republic of China: Economic and Social Commission for Asia and the Pacific, Mineral Resources Development Series No. 52, pp. 13– 21, United Nations, New York. SUPPORTING INFORMATION Supplementary data are available at GJI online. Table S1. The seismic networks used in this study. The number of stations in respective networks and their data collection periods for this study as well as their initial sampling frequency are also noted. Table S2. The events used to study frequency-dependent Lg attenuation in NE China. Table S3. The number of ray paths used in this study at different frequencies. Highest number of ray paths was found at a central frequency of 1.0 Hz. The number of ray paths decreases rapidly at 3.0 Hz frequency. Figure S1. Moho thickens map of NE China (adapted from Tao et al. 2014). Small squares represent seismic stations used by Tao et al. (2010). Most of Songliao Basin and parts of the Sino-Korean Craton show thinner crust (<32 km) than the rest of NE China. Abbreviations on the maps are: ELB, Erlian Basin; GXA, Great Xing’an range; HLB, Hailar Basin; SLB, Songliao Basin; ZGC, Zhangguangcai Range; SJB, Sanjiang Basin; JMS, Jiamusi Massif; SKC, Sino-Korean Craton. Figure S2.S-wave velocity model of NE China (adopted from Li et al. 2016). (a) S wave velocity inverted using both phase velocity and Z/H ratio data at a depth of 1 km. (b) Only using phase velocity data at a depth of 1 km. (c–f) Jointly inverted 3-D S wave velocity models at depths of 4, 6, 7 and 24 km. At shallower depths (≤ 6 km), sedimentary basins show low velocity values compared to mountainous regions. Figure S3. Frequency-dependent Lg Q in NE China at frequencies of (a) 0.5 Hz, (b) 1.0 Hz and (c) 2.0 Hz (adapted from Zhao et al. 2010). The Q values in the area are about 200–300 less than our model. Please note: Oxford University Press is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the paper. © The Author(s) 2017. Published by Oxford University Press on behalf of The Royal Astronomical Society. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Geophysical Journal International Oxford University Press

# Frequency-dependent Lg attenuation in Northeast China and its implications

, Volume 212 (3) – Mar 1, 2018
12 pages

/lp/ou_press/frequency-dependent-lg-attenuation-in-northeast-china-and-its-wkj2d3Dplf
Publisher
The Royal Astronomical Society
ISSN
0956-540X
eISSN
1365-246X
D.O.I.
10.1093/gji/ggx522
Publisher site
See Article on Publisher Site

### Abstract

Summary Lg attenuation and site responses in Northeast (NE) China are estimated by implementing the Reverse Two Station Method (RTSM) at narrow band central frequencies of 0.5, 1.0, 2.0 and 3.0 Hz using 453 earthquakes recorded by 201 seismic stations deployed in the region from 1995 to 2013. The RTSM has the advantage of removing source and site effects without requiring a priori models. Tomographic images produced at a resolution of 2° × 2° at all frequencies exhibit a high degree of lateral variation of Lg attenuation in NE China. The Great Xing’an, Lesser Xing’an and Songen-Zhangguangcai Ranges show Q values above 400 at all frequencies. At central frequencies of 0.5 and 1 Hz, Sanjiang Basin Songliao Basin, Erlian Basin and Hailar Basin consistently show Q values lower than 400. Holocene and Pleistocene volcanoes, including the Wudalianchi and Jingpuhi volcanic fields, also appear as regions of low Q (<400) at 0.5 and 1 Hz and at higher frequencies the effects of volcanoes diminish. At high frequencies (≥2 Hz), the sedimentary basins show Q values higher than 400 and overall Q values increase with frequency in NE China, thus obeying a power-law frequency dependence. A linear regression of frequencies in the range of 0.25–4.5 Hz results in the parameters describing the power-law frequency dependence of the region, with an average Q0 value of 428 and a frequency-dependent factor (η), describing the strength of dependence, of 0.68. Overall, Lg attenuation in NE China appears to be due to thick late Cretaceous sediments, Holocene and Pleistocene volcanism, moderate to high heat flow, partial melts and variation in the thickness of the crustal wave guide. The site responses calculated at 0.5, 1.0, 2.0 and 3 Hz show a high degree of lateral variation as well as variation with frequency in NE China. At 0.5 and 1.0 Hz, the Great Xing’an and Lesser Xing’an Ranges show deamplification while the Hailar, Erlian and southern Songliao basins and the Songen-Zhangguangcai Range show amplification. At higher frequencies (≥2 Hz), the pattern nearly reverses, with deamplification east of the Songliao Basin and amplification to the west of the Basin. The deamplification observed east of the Songliao Basin could be caused by basaltic lava flows resulting from Cenozoic volcanism in the region. Asia, Guided Waves, Seismic attenuation, Seismic tomography, Site effects 1 INTRODUCTION The tomographic mapping of crustal attenuation is an important method to understand regional crustal structure. Crustal attenuation is especially sensitive to sediments, fluids, partial melt and sudden variation in crustal thickness. Furthermore, its calibration is important for nuclear monitoring purposes and magnitude determination. Here, we examine the attenuation of the Lg wave in Northeast (NE) China using data from the NorthEast China Extended SeiSmic (NECESS) array experiment and other seismic experiments conducted in the region from 1995 to 2014. Lg is the most prominent high frequency wave that can be observed at regional distances and can be modelled as a superposition of multiple S wave reverberations and SV to P and P to SV conversions within the crustal waveguide (Bouchon 1982; Herrmann & Kijko 1983; Olsen et al. 1983; Campillo et al.1984; Kennett 1986; Ou & Herrmann 1990) or as the sum of higher mode surface waves (Oliver et al. 1955; Knopoff et al. 1973; Kennett 1984; Bostock & Kennett 1990). The average group velocity window of Lg ranges from 3.0 to 3.6 km s−1 and most of its energy lies within the 0.2 to 10 Hz frequency band in NE China. The efficient propagation of the Lg wave is highly dependent on the thickness of the crustal waveguide. In a simplified layer over a half space crustal model, the maximum number of viable modes necessary to generate Lg is proportional to the depth of the wave guide (Aki & Richards 1980; Zhang & Lay 1995). High frequency (>1 Hz) Lg ground motion at long distances (>200 km) is caused by supercritically reflected S waves bouncing back and forth between the crustal waveguide (e.g. Bouchon 1982; Herrmann 1985; Kennett 1986) as well as from supercritical reflections at internal interfaces (Burger et al. 1987). Using synthetic seismograms generated from ray theory, main Lg ground motion at regional distances has been shown to be caused by supercritically reflected S waves confined within the waveguide between the free surface and crustal interfaces below the source (Ou & Herrmann 1990). Thus, the Lg wave propagates efficiently over continental paths, having been observed at distances up to 30° in the Canadian Shield (Hasegawa 1985), which have sufficient crustal thickness. In contrast, very thin oceanic crust (<10 km) is unable to maintain these modes (e.g. Press & Ewing 1952; Knopoff et al. 1979; Kennett 1986; Zhang & Lay 1995), resulting in the complete blockage of Lg in oceanic crust (Cao & Muirhead 1993; Zhang & Lay 1995). This dependence of the modes on crustal thickness also results in the Lg wave quickly losing energy while propagating through areas that undergo rapid changes in crustal thickness, such as the transition that occurs beneath Tibet (e.g. McNamara et al. 1996; Xie 2002; Fan & Lay 2003). Lg attenuation has been shown to be sensitive to crustal attenuation structure (e.g. Baqer & Mitchell 1998), intrinsic properties of the crustal wave guide such as crustal temperature (e.g. Frankel 1991), high heat flow (e.g. Aleqabi & Wysession 2006), fluid content (e.g. Mitchell 1995), the presence of partial melts (e.g. Xie et al. 2004), unconsolidated sediments (e.g. Mitchell & Hwang 1987), crustal deformation (e.g. Pasyanos et al. 2009) and the age of the last crustal deformation event (e.g. Mitchell et al. 1997). NE China is a region of high geopolitical and tectonic interest. Only a few studies of regional seismic wave attenuation have been conducted in this region (Rapine & Ni 2003; Zhao et al. 2010) prior to the deployment of the NECESS array in 2009. Zhao et al. (2010) applied a single-station, single-event method to simultaneously invert for frequency-dependent Lg Q and the source function in NE China using 20 stations and 125 events. They found the area has an average Q0 (Q at 1 Hz) value of 414 and observed that the sedimentary basins have lower Q values while granitic mountain ranges were found to have higher Q values. Ranasinghe et al. (2015) used 78 crustal earthquakes recorded by the 127 station NECESS array to study Lg attenuation at 1 Hz in the region using the Reverse Two Station Method (RTSM). They observed that regions with little to no sediment (<1 km) tended to be characterized by high Q values (>800) while regions with moderately high heat flow (>70 mW m−2) tended to be characterized by low Q values (<800). Furthermore, they concluded Holocene volcanoes have a greater effect on Lg attenuation than Pleistocene volcanoes in the region. In an update of our prior study, we discuss results obtained from the RTSM for NE China, including estimates of frequency-dependent Q, and site responses. In this study, we address anomalies observed in the Lg attenuation tomography of NE China and their relationship to geology, sediment thickness, crustal thickness, and heat flow using a larger data set than our prior study and we also study frequency-dependent Lg attenuation and site responses. Comparisons with previous studies in the region will also be made. Finally, we will discuss the frequency dependence of Lg attenuation and the site responses observed in NE China. Our results in NE China confirms the frequency dependence of Lg attenuation that has been observed in other parts of the world (e.g. Chun et al.1987; Erickson et al.2004), which could be due to the effects of transitioning from a scattering-dominated regime to an intrinsic-dominated attenuation regime (Dainty 1981). 2 GEOLOGY AND TECTONICS OF NE CHINA Present day NE China is a geologically and structurally complex region composed of deep, organic-rich sedimentary basins, mountain ranges, and Pleistocene and Holocene volcanoes (Fig. 1). The Songliao, Hailar, Erlian, Sanjiang and Bohai Basins represent the major sedimentary basins in the region. The Great Xing’an, Lesser Xing’an and Songen-Zhangguangcai Ranges are located between the basins and they are mostly composed of granitic and volcanic rocks. Changbaishan, a Holocene volcanic complex located on the border between China and North Korea, is formed from a subduction-induced upwelling (Tang et al. 2014). Other Holocene volcanic fields include the Wudalianchi, Jingpuhi and Longgan volcanic complexes (Fig. 1). Figure 1. View largeDownload slide A geological map of northeast China, dashed black line represents the boundaries of the tomographic images. Black lines are active faults and dashed black lines are Late Triassic sutures. Red triangles represent volcanoes of Holocene age and open triangles represent volcanoes of Pleistocene age. Abbreviations on the maps are: BHB, Bohai Basin; CBV, Changbaishan Volcanic Field; DMF, Dunhua-Mishan Fault; EB, Erlian Basin; GB, Gobi Basin; GXR, Great Xing’an Range; HB, Hailar Basin; JMS, Jiayin-Mudanjiang Suture; JPV, Jingpuhi Volcanic Field; JYF, Jiamusi-Yitong Fault; LGV, Longgan Group of Volcanoes; LXR, Lesser Xing’an Range; NF, Nenjiang Fault; SLB, Songliao Basin; SNB, Sanjiang Basin; SS, Suolon Suture; SZR, Songen-Zhangguangcai Range; UBB, Upper Bureya Basin; WVF, Wudalianchi Volcanic Field; YTB, Yitong Basin; ZBB, Zeya–Bureya Basin. Figure 1. View largeDownload slide A geological map of northeast China, dashed black line represents the boundaries of the tomographic images. Black lines are active faults and dashed black lines are Late Triassic sutures. Red triangles represent volcanoes of Holocene age and open triangles represent volcanoes of Pleistocene age. Abbreviations on the maps are: BHB, Bohai Basin; CBV, Changbaishan Volcanic Field; DMF, Dunhua-Mishan Fault; EB, Erlian Basin; GB, Gobi Basin; GXR, Great Xing’an Range; HB, Hailar Basin; JMS, Jiayin-Mudanjiang Suture; JPV, Jingpuhi Volcanic Field; JYF, Jiamusi-Yitong Fault; LGV, Longgan Group of Volcanoes; LXR, Lesser Xing’an Range; NF, Nenjiang Fault; SLB, Songliao Basin; SNB, Sanjiang Basin; SS, Suolon Suture; SZR, Songen-Zhangguangcai Range; UBB, Upper Bureya Basin; WVF, Wudalianchi Volcanic Field; YTB, Yitong Basin; ZBB, Zeya–Bureya Basin. The Songliao Basin is composed of a series of interconnected half-grabens characterized by a thin crust (25–35 km) and relatively high heat flow (70 mWm−2). The basement of the Songliao Basin is cut by a system of Jurassic rifts, which are filled with volcanic rocks interlayered with Late Jurassic coal-bearing continental clastics (Zhou et al. 1985; Ulmishek 1992). In the Middle Cretaceous, during the early sag development stages of the Songliao Basin, sediment deposition was mainly confined to rift structures (Ulmishek 1992). In the Late Cretaceous, sediments were deposited in a large area extending up to 200 000 km2 from the centre. During this time, the deposition environment changed from an anoxic deepwater environment to a shallow oxic environment. Thus, shales dominate the central part of the basin while deltaic sandstones dominate the northern, southern and western margins (Ulmishek 1992). During the Cenozoic, NE China was characterized by extensive alkaline volcanism, the subduction of the Pacific plate underneath NE China, and the opening of the Sea of Japan. Volcanism in NE China was initially confined to the interior of the Songliao Basin and later (∼ 28 Ma) migrated to the flanks (Liu et al. 2001). Volcanic activity peaked in the Neogene along the Tan-Lu fault (TLF) and the Jiamusi-Yitong Fault (JYF; Liu et al. 2001) east of the Songliao Basin. Present day volcanism is dominated by the Changbaishan, Jingpuhi, and Wudalianchi volcanic complexes. Northeast China is currently undergoing east-west compression from the end of the opening of the Sea of Japan and the forces exerted by the subduction of the Pacific plate (Sagiya et al. 2000; Jin et al. 2007). 3 DATA PROCESSING The seismic waveform data used for this study were collected from 201 seismic stations operated in NE China and surrounding regions from January 1995 to December 2013 (Fig. 2). We used stations from several networks (Supporting Information Table S1), including: (1) 127 temporary broad-band stations from the NECESS Array (2011–2013), (2) 15 broad-band stations from the Incorporated Research Institute for Seismology (IRIS) Global Seismic Network (GSN) (1995–2013), (3) 19 broad-band stations from the NE China seismic experiment (1998–1999), (4) 26 broad-band stations from the natural and man-induced seismicity in the Yanquing-Hualia Basin and the Heicheng Area, China experiment (2002–2008), (5) 12 broad-band stations from the China National Seismic Network (2009–2013), (6) 1 broad-band station from the Korean Seismic Network (2013–2013), and (7) 1 broad-band station from the Japan meteorological agency seismic network (2007–2013). Figure 2. View largeDownload slide A map of the study region showing topography and the locations of all the crustal earthquakes used to study Lg attenuation. Magenta, black, red and purple circles represent the 453 earthquakes used to study Lg attenuation. The blue triangles represent the locations of the Northeast China Extended Seismic (NECESS) Array stations. The inverted magenta triangles represent the location of the Global Seismic Network (GSN) stations used in this study. The red stars represent the location of China National Seismic network stations. The purple diamonds represent the location of the NE China Seismic Experiment stations. The cyan stars represent the locations of the Natural & Man-Induced Seismicity in the Yanquing-Hualia Basin and the Heicheng Area, China experiment seismic stations. Finally, the red and cyan diamonds represent Korean Seismic Network and Japanese Metrological Agency seismic stations, respectively. Figure 2. View largeDownload slide A map of the study region showing topography and the locations of all the crustal earthquakes used to study Lg attenuation. Magenta, black, red and purple circles represent the 453 earthquakes used to study Lg attenuation. The blue triangles represent the locations of the Northeast China Extended Seismic (NECESS) Array stations. The inverted magenta triangles represent the location of the Global Seismic Network (GSN) stations used in this study. The red stars represent the location of China National Seismic network stations. The purple diamonds represent the location of the NE China Seismic Experiment stations. The cyan stars represent the locations of the Natural & Man-Induced Seismicity in the Yanquing-Hualia Basin and the Heicheng Area, China experiment seismic stations. Finally, the red and cyan diamonds represent Korean Seismic Network and Japanese Metrological Agency seismic stations, respectively. The sampling rate of the waveform data gathered by stations other than the NECESS Array network were adjusted to match the 40 samples s−1 rate of the NECESS Array stations prior to further processing. A list of the stations used in this study and their initial sampling frequency is included in the supporting material of this paper. The preliminary earthquakes chosen for this study range in magnitude from 2.5 to 7.0 (magnitude types are given in Supporting Information Table S2) and event depths were restricted to a maximum depth of 40 km, while epicentral distances were limited to a minimum of 2.5° and a maximum of 20°. A choice of focal depth of 40 km would make earthquakes more likely to be crustal in nature. The epicentral distance limits are in place because Lg does not develop prior to a distance of at least 2.5° while beyond 20° the Lg phase can no longer be distinguished from the pre-event noise in NE China. The waveform data were first processed by temporarily filtering the vertical component seismograms with a two-pass, three-pole recursive Butterworth filter with corners at 0.2 and 5 Hz for visualization purposes. The Pn arrival time was then picked to define the 20 s long pre-Pn noise window, with the end of the window starting at the pick. The Lg group velocity window was initially set to range from 3.6 km s−1 to 3.0 km s−1, as a visual guide to pick the onset of the Lg arrival. Then, the beginning of the Lg group velocity window was visually picked on each seismogram. If the characteristic Lg waveforms were absent, the seismograms were removed from further processing. After visually picking Pn and Lg arrival times, we chose 453 earthquakes (Fig. 2) with at least 3 clear Lg waveforms for an event (Supporting Information Table S2). The individual spectra were calculated by applying a Fourier transform with a 20 per cent cosine taper to the chosen pre-Pn noise window and Lg windows to avoid spectral leakage (Xie 1998). The signal-to-noise ratio (SNR) for Lg was calculated by dividing the respective Root Mean Square (RMS) of the Lg spectra by the RMS of the pre-Pn noise spectra in the frequency domain. Any records with an average SNR below 2.0 were removed from further processing. Finally, the instrument responses were removed from both the noise and signal spectra using instrument response information obtained from the IRIS Data Management Center (DMC). 4 METHOD 4.1 Amplitude parametrization The spectral amplitude of a seismic wave can be defined as   $$A\left( {f,d} \right) = S\left( f \right)I\!\left( f \right)E\!\left( f \right)G\!\left( d \right){\rm{exp}}\left( {\frac{{ - \pi fd}}{{vQ\left( f \right)}}} \right),$$ (1)where A(f, d) is the amplitude at a spectral frequency of f measured a given distance d from the source, S( f ) is the source response, I( f ) is the instrument response, E( f ) is the site response, G(d) is the geometrical spreading, $${\rm{exp}}\, ( {\frac{{ - \pi fd}}{{vQ( f )}}} )$$ is the frequency-dependent attenuation term, Q( f ) is the frequency-dependent quality factor, and v is the assumed average group velocity of 3.5 km s−1 for Lg. The geometric term for Lg is assumed to be in the form of   $$G\left( d \right) = {G_0}\ {d^{ - m}}$$ (2)where G0 is a constant (Chun et al.1987; Yang 2002; Bao et al.2011) and m is the geometric spreading coefficient. In this study, we implement the RTSM (Chun et al. 1987) to solve for the frequency-dependent quality factor Q(f), which describes Lg attenuation in NE China. Rather than invert for the geometric spreading coefficient m (Ranasinghe et al.2015), we use the generally accepted value of 0.5 (e.g. Yang 2002) in order to be consistent with previous Lg attenuation tomographic studies in the region (Zhao et al. 2010) and elsewhere (e.g. Xie 2002). 4.2 RTSM In the RTSM (Figs 3a and b), spectral amplitudes recorded at two stations (i and j) from two events (a and b) that lie along a great circle arc are divided out to remove the source response, site response and G0:   \begin{eqnarray} &&{\frac{{{A_{ai}}\left( {f,d} \right)}}{{{A_{aj}}\left( {f,d} \right)}}\ \frac{{{A_{bj}}\left( {f,d} \right)}}{{{A_{bi}}\left( {f,d} \right)}} = {\left( {\frac{{{d_{ai}}\ {d_{bj}}}}{{{d_{aj}}{d_{bi}}}}} \right)^{ - m}} } \nonumber\\ &&{\quad \times \, {\rm{exp}}\left[ {\frac{{\pi f}}{{vQ\left( f \right)}}\left( {{d_{aj}} - {d_{ai}} - {d_{bj}} + {d_{bi}}} \right)} \right]} \end{eqnarray} (3)where Aai and Aaj are the spectral amplitudes recorded by stations i and j respectively for the event a. Similarly, Abi and Abj are spectral amplitudes recorded by stations i and j for the event b. dai, daj, dbi and dbj are the corresponding epicentral distances as shown in Figs 3(a) and (b). In the RTSM we allow for a maximum variation of 15° from the great circle arc between the stations and the events, as shown in Fig. 3(b) (Der et al. 1984). The back azimuthal angle is also restricted to a maximum of 15° in order avoid possible obliqueness in longer ray paths. The path Q values are inverted following Xie & Mitchell (1990) to obtain the interstation Q values. Lg attenuation has been shown to follow a power law (e.g. Chun et al.1987; Erickson et al.2004) and it is defined as   $$Q\left( f \right) = {Q_0}{\left( {\frac{f}{{{f_0}}}} \right)^\eta }$$ (4)where, Q0 is the reference Q value at 1 Hz,  f0 is a reference frequency of 1 Hz and η is the frequency-dependent factor. The η represent variation in Q value with the frequency. Frequency-dependent pathQ values for the RTSM can be given as   $$\ln \left( {\frac{1}{{Q\left( f \right)}}} \right) = \ln \left( {\frac{1}{{{Q_0}}}} \right) + \ \eta {\rm{\ ln}}\left( {\frac{{{f_0}}}{f}} \right).$$ (5) Figure 3. View largeDownload slide An illustration of the Reverse Two Station Method (RTSM) configuration. The red stars a and b represent earthquakes and the black triangles i and j represent seismic stations. (a) The ideal geometry for the RTSM configuration. (b) The actual geometry for the RTSM configuration. The distance between stations i and j is represented by dij. The epicentral distances between the earthquakes a and b and the stations i and j are represented as dai, daj, dbi and dbj, respectively. θa and θb are the subtended angles between earthquakes a and b for stations i and j, respectively. Figure 3. View largeDownload slide An illustration of the Reverse Two Station Method (RTSM) configuration. The red stars a and b represent earthquakes and the black triangles i and j represent seismic stations. (a) The ideal geometry for the RTSM configuration. (b) The actual geometry for the RTSM configuration. The distance between stations i and j is represented by dij. The epicentral distances between the earthquakes a and b and the stations i and j are represented as dai, daj, dbi and dbj, respectively. θa and θb are the subtended angles between earthquakes a and b for stations i and j, respectively. The η and Q0 values for the RTSM are found using the intercept and the slope of the linear regression of eq. (5). 4.3 Site response The RTSM can also be used to calculate the relative site responses of individual stations (e.g. Bao et al. 2012; Gallegos et al.2017). These site responses are found by multiplying the amplitude ratios rather than dividing them as in eq. (3), resulting in eq. (6) below   \begin{eqnarray} \ln \ \left( {\frac{{{E_{i}}\left( f \right)}}{{{E_j}\left( f \right)}}} \right) &=& \frac{{{d_{aj}} - {d_{ai}}}}{{{d_{aj}} + {d_{bi}} - {d_{ai}} - {d_{bj}}}}\ \ln \left( {\frac{{{A_{ai}}\left( {f,d} \right)d_{ai}^m}}{{{A_{aj}}\left( {f,d} \right)d_{aj}^m}}} \right) \nonumber\\ &&+ \ \frac{{{d_{bi}} - {d_{bj}}}}{{{d_{aj}} + {d_{bi}} - {d_{ai}} - {d_{bj}}}}\ln \left( {\frac{{{A_{bi}}\left( {f,d} \right)d_{bi}^m}}{{{A_{bj}}\left( {f,d} \right)d_{bj}^m}}} \right). \nonumber\\ \end{eqnarray} (6)where Ei and Ej are the site responses of station i and  j, respectively. Eq. (6) can be formulated as a least squares problem. As the first step in calculating the site responses, eq. (6) is inverted using the LSQR algorithm (Paige & Saunders 1982). The site responses are calculated relative to a set of reference stations. These reference stations were varied to test the robustness of the model. Stations that had sampled a large number of ray paths had robust solutions that did not vary with the choice of reference stations. Thus, we simply chose the reference stations with the highest number of ray paths (NE21, NE23, NE62, NE76, NE77, NE82 and NE91). 5 RESULTS AND DISCUSSION 5.1 Lg tomography Utilizing RTSM, Lg Q tomography images (Fig. 4) with 2° × 2° resolution (Fig. 5) were produced at frequency bands of 0.25–0.75 Hz, 0.5–1.5 Hz, 1.0–3.0 Hz and 1.5–4.5 Hz with central frequencies of 0.5, 1.0, 2.0 and 3.0 Hz, respectively. The tomography and resolution maps were restricted to only include the area ranging from 42.5° to 48° latitude and from 116° to 135° longitude to minimize the edge effects from areas with low (<10) ray-path density (Fig. 6). We also did not include Lg path Q values less than 15 and greater than 2000 in the inversion, considering them outliers. A Q error ratio is also calculated by dividing the path Q value by its path Q error, whereas path Q error is one standard deviation of the best fit between the path Q values and frequency in each frequency spectra. Path Q error ratios greater than one were also removed from the inversion. We used a damping value of 0.75 and a Gaussian smoothing to keep the inversion stable and avoid having the edge effects bleed into the tomography images. Figure 4. View largeDownload slide Tomographic maps of Lg Q in NE China created using the RTSM at central frequencies of: (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz. The Q value increases with increasing frequency. Figure 4. View largeDownload slide Tomographic maps of Lg Q in NE China created using the RTSM at central frequencies of: (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz. The Q value increases with increasing frequency. Figure 5. View largeDownload slide 2° × 2° RTSM Lg Q resolution tests using a damping value of 0.15 and 15 per cent normally distributed noise at central frequencies of (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz. Lg attenuation is well-resolved at all frequencies with the exception of smearing at the edges. The Changbaishan and Wudalianchi volcanic fields are poorly resolved. Figure 5. View largeDownload slide 2° × 2° RTSM Lg Q resolution tests using a damping value of 0.15 and 15 per cent normally distributed noise at central frequencies of (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz. Lg attenuation is well-resolved at all frequencies with the exception of smearing at the edges. The Changbaishan and Wudalianchi volcanic fields are poorly resolved. Figure 6. View largeDownload slide RTSM Lg ray-path density at central frequencies of (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz using 3018, 4539, 4032 and 2724 ray paths, respectively. The best ray-path coverage is observed in the central part of NE China. A lack of ray-path coverage can be observed at the edges, particularly in the Changbaishan region and south of the Suolon Suture. Figure 6. View largeDownload slide RTSM Lg ray-path density at central frequencies of (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz using 3018, 4539, 4032 and 2724 ray paths, respectively. The best ray-path coverage is observed in the central part of NE China. A lack of ray-path coverage can be observed at the edges, particularly in the Changbaishan region and south of the Suolon Suture. The tomography images show increasing average Q values of 281, 411, 684 and 931 at 0.5, 1.0, 2.0 and 3.0 Hz, respectively. Uncertainties of the models were calculated by inverting path Q error values in the same fashion as the path Q values (Fig. 7). The best ray-path density is obtained in the Songliao Basin and the Great Xing’an Range (Fig. 6). Although a larger dataset was used in this study compared to Ranasinghe et al. (2015), it did not result in a substantial increase in the area of the ray-path coverage (Fig. 6) because of the restrictive RTSM geometry. However, the larger dataset did result in an overall increase in the number of ray paths used in this study (Supporting Information Table S3). The Songliao, Sanjiang, Hailar and Erlian Basins show low (< 400) Lg Q values at 0.5 Hz. The central part of the Songliao Basin, which is dominated by thick shale deposits, exhibits lower Q values compared to the edges of the basin, which are dominated by thin sand stone deposits (Ulmishek 1992). In a laboratory experiment, Winkler and Nur (1979) demonstrated that Q values increase as pore fluid is flushed out of a sand stone by increasing pressure. Therefore, in more compact sediments, the Q value is expected to be higher than in uncompacted, fluid-filled sediments (Mitchell & Hwang 1987). At 0.5 Hz, Sanjiang, Hailar and Erlian basins also show low Q values due to unconsolidated sediments. High attenuation experienced in sedimentary basins could also be due to fluids (Mitchell and Hwang 1987) and focusing and defocusing effects of the basins. The North and central Great Xing’an Ranges show high (>800) Q values and the southern Great Xing’an Range shows moderately high (400 to 800) Q values. The Lesser Xing’an Range and Songen-Zhangguangcai Ranges show moderately high Q values. Overall at 0.5 Hz, sedimentary basins with unconsolidated sediments and the Wudalianchi Volcanic Field show high attenuation and granitic mountain ranges show low attenuation. The results are similar at 1 Hz frequency, but there is reduced attenuation beneath the Greater Xing’an Range and along the Nenjiang Fault and increased attenuation in the east beneath the Lesser Xing’an Range and the southeastern Jiamusi-Yitong and Dunhua-Mishan Faults. Previous Lg attenuation studies have also indicated regions with active faulting, fluid filled fractures and suture zones tend to show both inefficient Lg propagation as well as high Lg attenuation (e.g. Crampin 1994; Mitchell 1995). Sandvol et al. (2001) observed inefficient Lg propagation in the Bitlis Suture and the Zagros fold and thrust belt in the Middle East. In another study in Tibet, Bao et al. (2012) found anisotropic 1/QLg is strongest in the northwestern Sonpan-Ganzi fold belt with high Q values running parallel to major strike-slip fault planes in the region. At 1 Hz, the Songen-Zhangguangcai and Great Xing’an Ranges show high Q values of 800–1000. Altogether, at 1 Hz sedimentary basins and the Wudalianchi and Jingpuhi Volcanic Fields show high attenuation while the Songen-Zhangguangcai and Great Xing’an Ranges show low attenuation. Figure 7. View largeDownload slide Lg Q errors calculated at (a) 0.5 Hz, (b) 1.0 Hz, (c) 2.0 Hz and (d) 3.0 Hz frequencies show most of the errors are less than 0.4 in areas with a high ray-path coverage. Errors are large at the edges of the tomographic images. Figure 7. View largeDownload slide Lg Q errors calculated at (a) 0.5 Hz, (b) 1.0 Hz, (c) 2.0 Hz and (d) 3.0 Hz frequencies show most of the errors are less than 0.4 in areas with a high ray-path coverage. Errors are large at the edges of the tomographic images. At higher frequencies, all the mountain ranges show high Q values and the Q values in the sedimentary basins increase compared to lower frequencies (≤1 Hz). The low Q values observed at low frequencies (≤1 Hz) correspond with thick unconsolidated sediments (Fig. 8), volcanic fields and fault lines found throughout NE China. At 2 Hz, the average Q has increased but Q variations are subdued apart from a high attenuation patch in the southeast corner and by 3 Hz the image is dominated by high Q values in the region. Figure 8. View largeDownload slide Lg site response maps of NE China overlaid on a sediment thickness (Laske & Masters 1997) map at central frequencies of (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz. The sizes of the triangles and the circles represent the values of the deamplification and the amplification, respectively. At central frequencies of 0.5 Hz and 1 Hz, most of the seismic stations in the basins show amplification. Additionally, at central frequencies of 2 Hz and 3 Hz, most of the seismic stations east of the Songliao except in the Sanjiang Basin show deamplification. Figure 8. View largeDownload slide Lg site response maps of NE China overlaid on a sediment thickness (Laske & Masters 1997) map at central frequencies of (a) 0.5 Hz, (b) 1.0 Hz, (c) 2 Hz and (d) 3 Hz. The sizes of the triangles and the circles represent the values of the deamplification and the amplification, respectively. At central frequencies of 0.5 Hz and 1 Hz, most of the seismic stations in the basins show amplification. Additionally, at central frequencies of 2 Hz and 3 Hz, most of the seismic stations east of the Songliao except in the Sanjiang Basin show deamplification. Both Pleistocene and Holocene volcanoes seem to have a large effect on Lg attenuation at lower (≤1 Hz) frequencies and at higher frequencies (≥2 Hz) both Pleistocene and Holocene volcanoes exhibit high Q values. Similar observations of Lg attenuation have been made in other parts of the world where Neogene and Cenozoic volcanism cause poor Lg propagation in middle east (e.g. Sandvol et al.2001) and high attenuation in the western U.S. (e.g. Phillips & Stead 2008; Gallegos et al. 2017). High Q values exhibited by the Quaternary volcanoes at higher frequencies could be due to less scattering experienced at higher frequencies. The Lg Q tomography maps created at different frequencies are consistent with previous models of Lg attenuation in NE China (Zhao et al. 2010) as well as with the geology in the region. These maps show an increase in Q values with frequency, that is, a power-law relationship (eq. 4) as predicted in other studies (e.g. Chun et al.1987; Erickson et al.2004; Zhao et al. 2010). We obtained a frequency-dependent factor (η value) of 0.68 and an average Q0 (Q at 1 Hz) value of 428 in NE China (Fig. 9). Figure 9. View largeDownload slide Plot of $$\ln ( {\frac{1}{Q}} )$$versus $$\ln ( {\frac{{{f_0}}}{f}} )$$ using the RTSM. Q0 and η values calculated from the linear fit for the RTSM are 0.68 and 428, respectively. Figure 9. View largeDownload slide Plot of $$\ln ( {\frac{1}{Q}} )$$versus $$\ln ( {\frac{{{f_0}}}{f}} )$$ using the RTSM. Q0 and η values calculated from the linear fit for the RTSM are 0.68 and 428, respectively. Comparing our Lg attenuation results with the geology of the region, we find that at low frequencies (≤1 Hz), sedimentary basins exhibit high attenuation (Q < 400) while mountain ranges show low attenuation (Q > 400). In the case of the basins, high attenuation is partially due to the presence of porous, fluid rich sediments and focusing/defocusing effects. These results are consistent with the previous Lg attenuation results at a central frequency of 1 Hz reported by Ranasinghe et al. (2015), where regions with thick sediments (>1 km) were characterized by low Q0 values (< 800). We also compared our Lg attenuation results to a sediment thickness map (Fig. 8) created using the global model made by Laske & Masters (1997). The map shows that the thickest sediments (∼7 km) are found in the central part of the Songliao Basin. The attenuation images produced at 0.5 Hz and 1.0 Hz show low attenuation (Q < 400) in areas with thick (>1 km) sediments. The attenuation images produced at 1.0 Hz show high attenuation in the Songliao, Hailar, Erlian and Sanjiang Basins. At central frequencies of 2 and 3 Hz, the thick sediments appear to cause very little attenuation, which could also be related to the increase of Q values in the basins with frequency. Overall, sediments are responsible for some of the high attenuation observed in the basins at lower frequencies. High heat flow has been also identified as a major cause of Lg attenuation in various parts of the world, such as the Tibetan Plateau (Xie 2002) and the Basin and Range province (Aleqabi & Wysession 2006). We compare our results with a heat flow study conducted by Wang and Cheng (2012) using a limited number of heat flow data values in the region. The authors found medium to high heat flow within the Songliao Basin, which corresponds to (70 mW m−2) an estimated average temperature of 600° C at the Moho beneath the basin. The high Lg attenuation found in the Songliao Basin at lower frequencies (≤1 Hz) could be partially due to the high heat flow, but there is insufficient heat flow data to conclusively state high heat flow is the main cause for the high attenuation observed in the Songliao Basin. Strong Lg attenuation and blockage has been observed in continental areas that undergo rapid changes in crustal thickness over relatively small distances, such as the Tibetan Plateau (Bàth 1954; Ni & Barazangi 1983; McNamara et al. 1996; Fan & Lay 2003), in the Bolivian Altiplano (Baumont et al. 1999) and in the European Alpine mountains (Campillo et al. 1993). We also compared our attenuation results with a crustal thickness map of NE China (Tao et al. 2014) created using the receiver functions (Supporting Information Fig. S1). The crustal thickness underneath Sanjiang Basin, Jingpuhi and Wudalianchi Volcanic Fields and the majority of Songliao Basin, with the exception of the northern portion, varies from ∼20 to 32 km. The crustal thickness in rest of NE China varies from ∼ 32 up to 45 km. We observed regions with thin crust, especially at the central frequency of 1 Hz, were found to show high attenuation but overall we did not observe a one to one relationship between crustal thickness and Lg attenuation. Upper crustal shear (S) wave maps produced at 1, 4, 6 and 7 km depths in NE China by Li et al. (2016) using joint inversion of Rayleigh wave ellipticity and phase velocity from the ambient noise and earthquake data collected from the NECESS Array show a broad low velocity zone in the northern part of the Songliao Basin and another low velocity zone near southwestern edge of the Basin as well as in Erlian, Hailar and Sanjiang Basins (Supporting Information Fig. S2). The low attenuation regions in the mountainous regions of the 1 Hz Lg attenuation image also seems to correspond with high velocity zones in the upper crustal shear wave maps created at 1, 4, 6, and 7 km depth. The low velocity zones at 1 and 4 km depth also corresponds with high Lg attenuations regions in our 0.5 and 1 Hz Lg attenuation images. Low S-wave velocities tend to correspond with low Lg Q values in the upper crust and it could be due to unconsolidated sediments causing low Q as well as low S-wave velocities in the upper crust. Zhao et al. (2010) used 1720 Lg spectra derived from 20 broad-band stations and 125 events to create Lg attenuation images in NE China at central frequencies of 0.5, 1.0 and 2 Hz using a single-event, single-station method (Supporting Information Fig. S3). They observed low Q values in the Bohai, Songliao, Hailar, Erlian, Sanjiang and Zeya-Buriya Basins and high Q values in the Great Xing’an, Lesser Xing’an and Songen-Zhangguangcai Ranges. Our RTSM results agree well with the broad regional features of Zhao et al. (2010). Being able to produce quite similar Lg attenuation tomography models implementing two different techniques and utilizing two different data sets can be taken as a strong measure of the robustness of both models. However, the Lg Q values calculated by Zhao et al. (2010) are frequently lower by ∼200–300 than those we calculated in our study. The difference between the Q values may have arisen from the fact that in our study we inverted for Lg Q−1, while Zhao et al. (2010) inverted for Q. By inverting for Lg Q−1 we avoid the potential bias introduced by averaging Lg Q values, which can be negatively affected by the presence of very large Lg Q values. Zhao et al. (2010) also observe an increase of Lg Q values with frequency, which is consistent with our study. Comparing with our previous study of Lg attenuation in NE China at a central frequency of 1 Hz (Ranasinghe et al.2015), our new models created at 0.5, 1.0, 2.0 and 3.0 Hz provide a broader view of Lg attenuation in NE China at different central frequencies. The 1 Hz images of the two studies show features such as sedimentary basins and Holocene volcanoes as high attenuation regions and granitic mountain ranges as low attenuation regions. The resulting Q values of the two studies cannot be compared directly, as the geometric spreading values used in the two studies are different from each other. 5.2 Frequency dependence of Lg In order to model the frequency dependence of Lg attenuation in NE China, the frequency dependence factor (η) and the Q0 values for the RTSM (eq. 5) were calculated plotting $${\log _e}(\frac{1}{Q})$$ against $${\log _e}( {\frac{{{f_0}}}{f}} )$$ using the average Q values found from the tomographic images produced at central frequencies of 0.5, 1.0, 1.5, 2.0, 2.5 and 3.0 Hz corresponding to narrow band frequencies of 0.25–0.75, 0.5–1.5, 0.75–2.25, 1.5–3.0, 1.75–3.75 and 2.0–4.5, respectively (Fig. 9). We obtained an η value of 0.68 and a Q0 of 428 for Lg attenuation in NE China using RTSM. Our results are also consistent with previous Lg attenuation studies by showing an increase in average Q value in the frequency range from 0.5 to 3.0 Hz (e.g.; Erickson et al.2004; Zhao et al. 2010). Frequency-dependent Lg attenuation can be interpreted based on the definition of apparent Q as a linear combination of scattering Q and intrinsic Q (Dainty 1981). Dainty (1981) put forth the idea that Lg attenuation is dominated by scattering Q at lower frequencies (≤1 Hz) and by intrinsic Q at higher frequencies (>1 Hz) by intrinsic Q. The intrinsic Q value stays roughly constant throughout the frequency range while the importance of scattering decreases with an increase in frequency. Under this assumption, the observed power law behaviour of frequency is a result of the contributions from both intrinsic Q and scattering Q. The attenuation maps created at 0.5 and 1 Hz are dominated by scattering attenuation and the maps created at 2 and 3 Hz are dominated by intrinsic attenuation. It should be noted that the true mechanisms behind scattering and intrinsic attenuation are still under debate. The explanation put forth by Dainty (1981) is just one possibility. 5.3 Site response for Lg Frequency-dependent Lg site responses were calculated using eq. (6) at central frequencies of 0.5, 1.0, 2.0 and 3.0 Hz (Fig. 8). At 0.5 Hz for Lg, seismic stations in the Great Xing’an and Lesser Xing’an Ranges consistently show deamplification (negative site response relative to mean) while stations in the Hailar and Sanjiang Basins as well as the southern Songliao Basin and Songen-Zhangguangcai Range show amplification (positive site response relative to mean). However, this pattern nearly reverses at higher frequencies, with deamplification to the east of the Songliao basin and amplification to the west now being observed. At 3 Hz the site responses have the largest effect. Many of the stations located over areas with the thinnest sediments (∼0.5–1.0 km) show the largest changes in amplification with frequency. The station log-gains have values from −0.5 to +0.5. For an average Q of 400 over a 200 km ray-path segment, we expect close to this value for the log-amplitude ratio. At central frequencies of 2 and 3 Hz, stations in the mountain ranges east of the Songliao Basin show consistent and dramatic deamplification. Data shows that signal exists up to 5 Hz and these stations include both permanent and temporary stations. Thus, the deamplification cannot be attributed to data quality or deployment issues and we conclude the effect is due to local site geology. The region does have numerous Cenozoic volcanoes, granitic rocks and basalt flows, and it also lies between Dunhua-Mishan Fault and Jiamusi-Yitong Fault. Phillips and Aki (1986) observed the strongest Lg coda site deamplification for the frequency range of 1.5–3.0 Hz in the granitic Gabilan Range and meta-sediment Franciscan basement sites in California. They also observe site amplification is inversely proportional to deposit age at sedimentary sites. They further observed moderate site amplification in Miocene and older Cenozoic sediment deposits. Thus, we conclude the strong deamplification observed east of the Songliao Basin at higher frequencies is partly due to Cenozoic volcanism. However, we also note that deamplification at these higher frequencies cannot be solely attributed to sedimentation or surface rock type. Lg site responses in the Erlian Basin, which is composed of half graben systems (Changsong et al. 2001), vary drastically between frequencies while most of the stations outside the Erlian Basin show high amplification at 2 Hz. It could be related to near surface features and focusing-defocusing effects from grabens and half-grabens present in the Basin. Akyol et al. (2013) also observed ground amplification in a horst-graben system in western Turkey due to focusing-defocusing effects. They also observed large seismic amplitude variations in seismic stations closer to the basin edges due to the complex nature of the wave propagation. Theoretically, it has also been shown that elastic focusing/defocusing effects in a 3-D Earth model with lateral variations in velocity and anelastic attenuation causes perturbations in the amplitude of surface waves (e.g. Zhou 2009; Ruan and Zhou 2012). Therefore, much of the amplification observed at the edges of the basin could be attributed to focusing and defocusing effects. 6 CONCLUSION The RTSM was used to measure lateral variations in Lg Q in NE China using an extensive data set collected from 1995 to 2013 in the region by various seismic networks. This method eliminates most of the effects of both the source and site terms. Lg Q maps were created applying the RTSM method at central frequencies of 0.5, 1.0, 2.0 and 3.0 Hz which can resolve 2° × 2° features, with the best resolution found in the Great Xing’an Range and Songliao Basin. Average Lg Q values obtained for NE China increase with frequency and we obtained a frequency-dependent factor (η) of 0.68 and an average Q0 (Q value at 1 Hz) of 428 for RTSM. A high degree of lateral variation in Lg attenuation was found in the region at all frequencies. At low frequencies (≤1 Hz), the south and eastern part of the Songliao Basin, as well as the Erlian, Hailar and Sanjiang Basins show high attenuation. The Great Xing’an Range, Lesser Xing’an Range and Songen-Zhangguangcai Range show low attenuation at low frequencies. At low frequencies, areas of high Lg attenuation were found to correspond to thick sediments and at higher frequencies ( ≥ 2 Hz) the correlation ceased to exist. Also at low frequencies, high attenuation corresponds with low shear velocity, high heat flow and Quaternary volcanism in the area. Regions with low attenuation correspond to mountain ranges and areas that have not experienced recent tectonic activity. The high attenuation observed at lower frequencies is more dominated by scattering processes rather than intrinsic attenuation. Lg site responses are found to vary laterally as well as with frequency throughout NE China. At all frequencies, the majority of seismic stations lying in the mountain ranges show deamplification. Amplifications in the basins vary widely with frequency. The observed seismic amplification in the basins could be related to focusing-defocusing effects. The deamplification observed east of the Songliao Basin at higher frequencies (≥2 Hz) could be related to Cenozoic volcanism in the region. ACKNOWLEDGEMENTS We would like to thank the NECESS Array (doi:10.7914/SN/YP_2009) team from China, Japan and the USA for collecting most of the data used in this study. We also used seismic data collected by the China National Seismic Network (doi:10.7914/SN/CB), the Global Seismograph Network (doi:10.7914/SN/IU, doi:10.7914/SN/II), the New China Digital Seismograph Network (doi:10.7914/SN/IC), the NE China Seismic Experiment, the Natural and Man-Induced Seismicity experiment in the Yanquing-Hualia Basin and the Heicheng Area, China (doi:10.7914/SN/XG_2002), and finally the Korean Seismic Network and Japan Metrological Agency seismic network. We would also like to thank the Incorporated Research Institutions for Seismology (IRIS) Data management System (DMC) for allowing us to access the data and promptly answering the queries regarding the data. Comments and suggestions from Jack Xie and another anonymous reviewer have been a great help in improving the manuscript. Finally, we would like to thank Paul Wessel and Walter H.F. Smith for the use of the Generic Mapping Tools (GMT) software, Lawrence Livermore for the use of their Seismic Analysis Code (SAC), the National Geophysical Data Center of NOAA for making the Etopo1 global relief map freely available and Fugro Robertson, Ltd for making the basin classification database freely available through AAPG. New Mexico State University is supported under National Science Foundation (NSF) EAR-0608629, U.S. Air Force Research Laboratory contract No. FA9453-12-C-0235 and Los Alamos National Security Subcontract No. 218475. REFERENCES Aki K., Richards P.G. 1980. Quantitative Seismology , 2nd edn, pp. 251– 253, W.H. Freeman and Co. Akyol N., Kurtulmuş T.Ö., Çamyildiz M., Güngör T., 2013. Spectral ratio estimates for site effects on the horst-graben system in West Turkey, Pure appl. Geophys. , 170( 12), 2107– 2125. https://doi.org/10.1007/s00024-013-0661-2 Google Scholar CrossRef Search ADS   Aleqabi G.I., Wysession M.E., 2006. QLg distribution in the basin and range province of the Western United States, Bull. seism. Soc. Am. , 96( 1), 348– 354. https://doi.org/10.1785/0120040086 Google Scholar CrossRef Search ADS   Bao X., Sandvol E., Ni J., Hearn T., Chen Y.J., Shen Y., 2011. High resolution regional seismic attenuation tomography in eastern Tibetan Plateau and adjacent regions, Geophys. Res. Lett. , 38( 16), L16304, doi:10.1029/2011GL048012. https://doi.org/10.1029/2011GL048012 Google Scholar CrossRef Search ADS   Bao X., Sandvol E., Chen Y.J., Ni J., Hearn T., Shen Y., 2012. Azimuthal anisotropy of Lg attenuation in eastern Tibetan Plateau. J. geophys. Res. , 117( B10), B10309, doi:10.1029/2012JB009255. https://doi.org/10.1029/2012JB009255 Google Scholar CrossRef Search ADS   Baqer S., Mitchell B., 1998. Regional variation of Lg coda Q in the continental United States and its relation to crustal structure and evolution, Pure appl. Geophys. , 153( 4), 613– 638. https://doi.org/10.1007/s000240050210 Google Scholar CrossRef Search ADS   Bàth M., 1954. The elastic waves Lg and Rg along Euro-Asiatic paths, Ark. Geofys. , 2, 295– 324. Baumont D., Paul A., Beck S., Zandt G., 1999. Strong crustal heterogeneityin the Bolivian Altiplano as suggested by attenuation of Lg waves, J. Geophys. Res. , 104( B9), 287– 305. Google Scholar CrossRef Search ADS   Bostock M.G., Kennett B.L.N., 1990. The effect of 3-D structure on Lg propagation patterns, Geophys. J. Int. , 101( 2), 355– 364. https://doi.org/10.1111/j.1365-246X.1990.tb06574.x Google Scholar CrossRef Search ADS   Bouchon M., 1982. The complete synthesis of seismic crustal phases at regional distances, J. geophys. Res. , 87( B3), 1735– 1741. https://doi.org/10.1029/JB087iB03p01735 Google Scholar CrossRef Search ADS   Burger R.W., Somerville P.G., Barker J.S., Herrmann R.B., Helmberger D.V., 1987. The effect of crustal structure on strong ground motion attenuation relations in eastern North America, Bull. seism. Soc. Am. , 77, 420– 439. Campillo M., Bouchon M., Massinon B., 1984. Theoretical study of the excitation, spectral characteristics and geometrical attenuation of regional seismic phases, Bull. seism. Soc. Am. , 74, 79– 90. Campillo M., Feignier B., Bouchon B., Be'thoux N., 1993. Attenuation of crustal waves across the Alpine range, J. geophys. Res. , 98( B2), 1987– 1996. https://doi.org/10.1029/92JB02357 Google Scholar CrossRef Search ADS   Cao S., Muirhead K.J., 1993. Finite difference modelling of Lg blockage, Geophys. J. Int. , 115( 1), 85– 96. https://doi.org/10.1111/j.1365-246X.1993.tb05590.x Google Scholar CrossRef Search ADS   Changsong L., Eriksson K., Sitian L., Yongxian W., Jianye R., Yanmei Z., 2001. Sequence architecture, depositional systems, and controls on development of lacustrine basin fills in part of the Erlian Basin, northeast China, Am. Assoc. Pet. Geol. Bull. , 85, 2017– 2043. Chun K.-Y., West G.F., Kokoski R.J., Samson C., 1987. A novel technique for measuring Lg attenuation results from eastern Canada between 1 to 10 Hz, Bull. seism. Soc. Am. , 77, 398– 419. Crampin S., 1994. The fracture criticality of crustal rocks, Geophys. J. Int. , 118( 2), 428– 438. https://doi.org/10.1111/j.1365-246X.1994.tb03974.x Google Scholar CrossRef Search ADS   Dainty A.M., 1981. A scattering model to explain seismic Q observations in the lithosphere between 1 and 30 Hz, Geophys. Res. Lett. , 8( 11), 1126– 1128. https://doi.org/10.1029/GL008i011p01126 Google Scholar CrossRef Search ADS   Der Z.A., Marshall M.E., O’Donnell A., McElfresh T.W., 1984. Spatial coherence structure and attenuation of the Lg phase, site effects, and interpretation of the Lg coda, Bull. seism. Soc. Am. , 74, 1125– 1148. Erickson D., McNamara D., Benz H., 2004. Frequency-dependent Lg Q within the continental United States, Bull. seism. Soc. Am. , 94( 5), 1630– 1643. https://doi.org/10.1785/012003218 Google Scholar CrossRef Search ADS   Fan G.W., Lay T., 2003. Strong Lg attenuation in the Tibetan Plateau, Bull. seism. Soc. Am. , 93( 5), 2264– 2272. https://doi.org/10.1785/0120030052 Google Scholar CrossRef Search ADS   Frankel A., 1991. Mechanisms of seismic attenuation in the crust: scattering and anelasticity in New York State, South Africa, and southern California, J. Geophys. Res. , 96, 6269– 6289. Google Scholar CrossRef Search ADS   Gallegos A., Ranasinghe N., Ni J., Sandvol E., 2017. Lg attenuation, frequency dependence and relative site response of the western United States as revealed by the EarthScope Transportable Array, Geophys. J. Int. , 209( 3), 1955– 1971. https://doi.org/10.1093/gji/ggx145 Google Scholar CrossRef Search ADS   Hasegawa H.S., 1985. Attenuation of Lg waves in the Canadian Shield, Bull. seism. Soc. Am. , 75, 1569– 1582. Herrmann R.B., 1985. An extension of random vibration theory estimates of strong ground motion to large distances, Bull. seism. Soc. Am. , 75, 1447– 1453. Herrmann R.B., Kijko A., 1983. Modeling some empirical vertical component Lg relations, Bull. seism. Soc. Am. , 73, 157– 171. Jin S., Park P.-H., Zhu W., 2007. Micro-plate tectonics and kinematics in Northeast Asia inferred from a dense set of GPS observations, Earth planet. Sci. Lett. , 257( 3–4), 486– 496. https://doi.org/10.1016/j.epsl.2007.03.011 Google Scholar CrossRef Search ADS   Kennett B.L.N., 1984. Guided wave propagation in laterally varying media—I. Theoretical development, Geophys. J. Int. , 79( 1), 235– 255. https://doi.org/10.1111/j.1365-246X.1984.tb02853.x Google Scholar CrossRef Search ADS   Kennett B.L.N., 1986. Lg Waves and structural boundaries, Bull. seism. Soc. Am. , 76, 1133– 1141. Knopoff L., Schwab F., Kausel E., 1973. Interpretation of Lg, Geophys. J. Int. , 33( 4), 389– 404. https://doi.org/10.1111/j.1365-246X.1973.tb02375.x Google Scholar CrossRef Search ADS   Knopoff L., Mitchel R.G., Kauself E.G., 1979. A search for the oceanic Lg Phase, Geophys. J. Int. , 56( 1), 211– 218. https://doi.org/10.1111/j.1365-246X.1979.tb04776.x Google Scholar CrossRef Search ADS   Laske G., Masters G., 1997. A global digital map of sediment thickness, EOS, Trans. Am. geophys. Un. , 78, F483. Li G., Chen H., Niu F., Guo Z., Yang Y., Xie J., 2016. Measurement of Rayleigh wave ellipticity and its application to the joint inversion of high-resolution S wave velocity structure beneath northeast China, J. geophys. Res. , 121( 2), 864– 880. https://doi.org/10.1002/2015JB012459 Google Scholar CrossRef Search ADS   Liu J., Han J., Fyfe W.S., 2001. Cenozoic episodic volcanism and continental rifting in northeast China and possible link to Japan Sea development as revealed from K-Ar geochronology, Tectonophysics , 339( 3–4), 385– 401. https://doi.org/10.1016/S0040-1951(01)00132-9 Google Scholar CrossRef Search ADS   McNamara D.E., Owens T.J., Walter W.R., 1996. Propagation characteristics of Lg across the Tibetan Plateau, Bull. seism. Soc. Am. , 86( 2), 457– 469. Mitchell B.J., 1995. Anelastic structure and evolution of the continental crust and upper mantle from seismic surface wave attenuation, Rev. Geophys. , 33( 4), 441– 462. https://doi.org/10.1029/95RG02074 Google Scholar CrossRef Search ADS   Mitchell B.J., Hwang H., 1987. Effect of low Q sediments and crustal Q on Lg Attenuation in the United States, Bull. seism. Soc. Am. , 77( 4), 1197– 1210. Mitchell B.J., Pan Y.P., Xie J., Cong L., 1997. Lg coda Q variation across Eurasia and its relation to crustal evolution, J. Geophys. Res. , 102( B10), 22 767– 22 779. Google Scholar CrossRef Search ADS   Ni J., Barazangi M., 1983. High-frequency seismic wave propagation beneath the Indian Shield, Himalayan Arc, Tibetan Plateau and surrounding regions: high uppermost mantle velocities and efficient Sn propagation beneath Tibet, Geophys. J. Int. , 72( 3), 665– 689. https://doi.org/10.1111/j.1365-246X.1983.tb02826.x Google Scholar CrossRef Search ADS   Oliver J., Ewing M., Press F., 1955. Crustal structure of the Arctic regions from the Lg Phase, Bull. geol. Soc. Am. , 66( 9), 1063– 1074. https://doi.org/10.1130/0016-7606(1955)66%5b1063:CSOTAR%5d2.0.CO;2 Google Scholar CrossRef Search ADS   Olsen K.H., Braile L.W., Stewart J.N., 1983. Modeling short-period crustal phases (p, Lg) for long-range refraction profiles, Phys. Earth planet. Inter. , 31( 4), 334– 347. https://doi.org/10.1016/0031-9201(83)90093-6 Google Scholar CrossRef Search ADS   Ou G.B., Herrmann R.B., 1990. A statistical model for ground motion produced by earthquakes at local and regional distances, Bull. seism. Soc. Am. , 80( 6A), 1397– 1417. Paige C.C., Saunders M.A., 1982. LSQR: An algorithm for sparse linear equations and sparse least squares, ACM Trans. Math. Softw. , 8( 1), 43– 71. https://doi.org/10.1145/355984.355989 Google Scholar CrossRef Search ADS   Pasyanos M.E., Matzel E.M., Walter W.R., Rodgers A.J., 2009. Broad-band Lg attenuation modelling in the Middle East, Geophys. J. Int. , 177( 3), 1166– 1176. https://doi.org/10.1111/j.1365-246X.2009.04128.x Google Scholar CrossRef Search ADS   Phillips W.S., Aki K., 1986. Site amplification of coda waves from local earthquakes in central California, Bull. seism. Soc. Am. , 76, 627– 648. Phillips W.S., Stead R.J., 2008. Attenuation of Lg in the western US using the USArray, Geophys. Res. Lett. , 35( 7), L07307, doi:10.1029/2007GL032926. https://doi.org/10.1029/2007GL032926 Google Scholar CrossRef Search ADS   Press F., Ewing M., 1952. Two slow surface waves across North America, Bull. seism. Soc. Am. , 42, 219– 228. Ranasinghe N.R. et al.  , 2015. Lg attenuation in northeast China using NECESSArray data, Geophys. J. Int. , 200( 1), 67– 76. https://doi.org/10.1093/gji/ggu375 Google Scholar CrossRef Search ADS   Rapine R.R., Ni J.F., 2003. Propagation Characteristics of Sn and Lg in Northeastern China and Mongolia, Bull. seism. Soc. Am. , 93( 2), 939– 945. https://doi.org/10.1785/0120010157 Google Scholar CrossRef Search ADS   Ruan Y., Zhou Y., 2012. The effects of 3-D anelasticity (Q) structure on surface wave amplitudes, Geophys. J. Int. , 189( 2), 967– 983. https://doi.org/10.1111/j.1365-246X.2011.05356.x Google Scholar CrossRef Search ADS   Sagiya T., Miyazaki S., Tada T., 2000. Continuous GPS array and present-day crustal deformation of Japan, Pure appl. Geophys. , 157, 2303– 2322. Sandvol E. et al.  , 2001. Tomographic imaging of Lg and Sn propagation in the Middle East, Pure appl. Geophys. , 158( 7), 1121– 1163. https://doi.org/10.1007/PL00001218 Google Scholar CrossRef Search ADS   Tang Y. et al.  , 2014. Changbaishan volcanism in northeast China linked to subduction-induced mantle upwelling, Nat. Geosci. , 7( 6), 470– 475. https://doi.org/10.1038/ngeo2166 Google Scholar CrossRef Search ADS   Tao K. et al.  , 2014. Crustal structure beneath NE China imaged by NECESSArray receiver function data, Earth planet. Sci. Lett. , 398, 48– 57. https://doi.org/10.1016/j.epsl.2014.04.043 Google Scholar CrossRef Search ADS   Ulmishek G., 1992. Geology and hydrocarbon resources of onshore basins in eastern China, U.S. Geological Survey open-file report; 93–4, pp. 48– 57, Denver, CO, USA. Wang Y., Cheng S.-H., 2012. Lithospheric thermal structure and rheology of the eastern China, J. Asian Earth Sci. , 47, 51– 63. https://doi.org/10.1016/j.jseaes.2011.11.022 Google Scholar CrossRef Search ADS   Winkler K., Nur A., 1979. Pore fluids and seismic attenuation in rocks, Geophys. Res. Lett. , 6( 1), 1– 4. https://doi.org/10.1029/GL006i001p00001 Google Scholar CrossRef Search ADS   Xie J., 1998. Spectral inversion of Lg from earthquakes: a modified method with application to the 1995, western Texas earthquake sequence, Bull. seism. Soc. Am. , 88, 1525– 1537. Xie J.K., 2002. Lg Q in the eastern Tibetan Plateau, Bull. seism. Soc. Am. , 92( 2), 871– 876. https://doi.org/10.1785/0120010154 Google Scholar CrossRef Search ADS   Xie J.K., Mitchell B.J., 1990. A back-projection method for imaging large-scale lateral variations of Lg coda Q with application to continental Africa, J. geophys. Res. , 100, 161– 181. Xie J., Gok R., Ni J., Aoki Y., 2004. Lateral variations of crustal seismic attenuation along the INDEPTH profiles in Tibet from Lg Q inversion, J. Geophys. Res. , 109, B10308, doi:10.1029/2004JB002988. Yang X., 2002. A numerical investigation of Lg geometrical spreading, Bull. seism. Soc. Am. , 92( 8), 3067– 3079. https://doi.org/10.1785/0120020046 Google Scholar CrossRef Search ADS   Zhang T.-R., Lay T., 1995. Why the Lg phase does not traverse oceanic crust, Bull. seism. Soc. Am. , 85( 6), 1665– 1678. Zhao L.-F., Xie X.-B., Wang W.-M., Zhang J.-H., Yao Z.-X., 2010. Seismic Lg-wave Q tomography in and around Northeast China, J. geophys. Res. , 115, B08307, doi:10.1029/2009JB007157. https://doi.org/10.1029/2009JB007157 Zhou Y., 2009. Surface-wave sensitivity to 3-D anelasticity, Geophys. J. Int. , 178, 1403– 1410. https://doi.org/10.1111/j.1365-246X.2009.04230.x Google Scholar CrossRef Search ADS   Zhou Z., Songyu Q., Changnian H., 1985. Songliao Basin, in ESCAP Atlas of Stratigraphy IV, Peoples's Republic of China: Economic and Social Commission for Asia and the Pacific, Mineral Resources Development Series No. 52, pp. 13– 21, United Nations, New York. SUPPORTING INFORMATION Supplementary data are available at GJI online. Table S1. The seismic networks used in this study. The number of stations in respective networks and their data collection periods for this study as well as their initial sampling frequency are also noted. Table S2. The events used to study frequency-dependent Lg attenuation in NE China. Table S3. The number of ray paths used in this study at different frequencies. Highest number of ray paths was found at a central frequency of 1.0 Hz. The number of ray paths decreases rapidly at 3.0 Hz frequency. Figure S1. Moho thickens map of NE China (adapted from Tao et al. 2014). Small squares represent seismic stations used by Tao et al. (2010). Most of Songliao Basin and parts of the Sino-Korean Craton show thinner crust (<32 km) than the rest of NE China. Abbreviations on the maps are: ELB, Erlian Basin; GXA, Great Xing’an range; HLB, Hailar Basin; SLB, Songliao Basin; ZGC, Zhangguangcai Range; SJB, Sanjiang Basin; JMS, Jiamusi Massif; SKC, Sino-Korean Craton. Figure S2.S-wave velocity model of NE China (adopted from Li et al. 2016). (a) S wave velocity inverted using both phase velocity and Z/H ratio data at a depth of 1 km. (b) Only using phase velocity data at a depth of 1 km. (c–f) Jointly inverted 3-D S wave velocity models at depths of 4, 6, 7 and 24 km. At shallower depths (≤ 6 km), sedimentary basins show low velocity values compared to mountainous regions. Figure S3. Frequency-dependent Lg Q in NE China at frequencies of (a) 0.5 Hz, (b) 1.0 Hz and (c) 2.0 Hz (adapted from Zhao et al. 2010). The Q values in the area are about 200–300 less than our model. Please note: Oxford University Press is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the paper. © The Author(s) 2017. Published by Oxford University Press on behalf of The Royal Astronomical Society.

### Journal

Geophysical Journal InternationalOxford University Press

Published: Mar 1, 2018

## You’re reading a free preview. Subscribe to read the entire article.

### DeepDyve is your personal research library

It’s your single place to instantly
that matters to you.

over 12 million articles from more than
10,000 peer-reviewed journals.

All for just $49/month ### Explore the DeepDyve Library ### Unlimited reading Read as many articles as you need. Full articles with original layout, charts and figures. Read online, from anywhere. ### Stay up to date Keep up with your field with Personalized Recommendations and Follow Journals to get automatic updates. ### Organize your research It’s easy to organize your research with our built-in tools. ### Your journals are on DeepDyve Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more. All the latest content is available, no embargo periods. ### Monthly Plan • Read unlimited articles • Personalized recommendations • No expiration • Print 20 pages per month • 20% off on PDF purchases • Organize your research • Get updates on your journals and topic searches$49/month

14-day Free Trial

Best Deal — 39% off

### Annual Plan

• All the features of the Professional Plan, but for 39% off!
• Billed annually
• No expiration
• For the normal price of 10 articles elsewhere, you get one full year of unlimited access to articles.

$588$360/year

billed annually

14-day Free Trial