We present new 3-D tomographic models of VP, VS and VP/VSratio anomalies in the upper mantle beneath EC and adjacent areas. This data was collected and interpreted with the goal of clarifying geodynamic processes that caused spatially variable event histories throughout Eastern China (EC) during Mesozoic to Cenozoic time. The tomographic images were constructed by inverting body wave travel-times recorded at ∼1300 stations within the upgraded China National Seismic Network, and 9 temporary arrays. Resolution tests for different depths and the featured velocity anomalies verify that the tomographic images capture the velocity heterogeneities in the upper mantle to depths of 700 km. The salient features of VP, VS and VP/VSratio anomalies can be clearly identified. These include strong multiscale heterogeneities occupying the upper mantle beneath EC and differences in the spatial scale of anomalies found beneath northern and southern areas of EC. These features demonstrate a degree of spatial variability in the geodynamic evolution of EC. We propose two mechanisms to explain these patterns. First, the western front of the subducted slab may have imparted greater horizontal compressional stress in areas where it impinged further eastward into EC. These areas would experience stronger convection and an altered stress regime in the upper mantle, creating significant thermal anomalies beneath the South China Block (SCB) relative to the eastern North China Craton (NCC). Second, differing thermal states and viscosities for the eastern NCC and the Cathaysia Block (CaB) resulted in differing responses to regional deformation. The Archean hinterland of the eastern NCC specifically has a colder thermal state and higher viscosity, and therefore exhibits only small-scale heterogeneities due to the effect of shear localization. The Neoproterozoic CaB has a relatively warm thermal state with lower viscosity, and thus deformed more continuously.
 As the eastern margin of the Eurasian plate, Eastern China (EC) records the geodynamic evolution associated with the ongoing convergence of Eurasia and the Paleo- Pacific and Philippine plates initiated during Mesozoic to Cenozoic time (Figure 1). Previous multidisciplinary studies have shown that the deformation of EC associated with plate convergence exhibits significant temporal and spatial variability that generally conforms to north, south, east and west boundaries of the region's major tectonic provinces [e.g., Ren et al., 2002; Wu et al., 2005; Niu, 2005; Wang et al., 2006; Li and Li, 2007; Sun et al., 2007; Yang et al., 2008; Yin, 2010; Zheng et al., 2012]. EC therefore provides an excellent natural laboratory for studying continental margin deformation in response to subduction. The geodynamic processes responsible for temporal and spatial patterns of deformation in this region have been examined and discussed at length, but without consensus on a consistent interpretation of the observations.
 This study capitalized on recent upgrades in the China National Seismic Network (CNSN), which provides adequate to relatively dense station coverage throughout EC. Data was also collected using portable stations across key tectonic units and boundaries. The combined VP, VS and VP/VSratio models described here are based on enhanced station coverage and thus provide novel, high-resolution information on the region's geodynamic response to subduction.
2. Geological Background
 We use the term “Eastern China” to describe the Chinese part of the eastern margin of the Eurasian plate that is tectonically linked to the subduction of the Pacific and Philippine Sea plates (Figure 1). For the sake of this study, EC is divided into three blocks from north to south: the Northeastern China Block, the North China Craton (NCC) and the South China Block (SCB). Along EC's eastern margin, the Pacific and Philippine Sea plate subduction zones extend beneath the Eurasian plate. The NE trending Xing'an-Mongolian orogenic belt welds the northern part of EC to the Siberian Craton. To the southwest, the NW-SE trending Longmenshan fault separates EC from the Tibetan Plateau.
2.1. Tectonic Origins of EC
 EC consists of a series of Phanerozoic orogenic belts welded together by Late Paleozoic to Early Mesozoic tectonic events. Prior to its amalgamation, EC could be divided into two separate tectonic provinces - the NCC and the SCB, both of which exhibit unique inherited properties. The NCC preserves continental rocks as old as 3.8 Ga [Liu et al., 1992]. Zhao et al. proposed that the NCC formed during Paleoproterozoic amalgamation of eastern and western Archean blocks. Basement units of the NCC are overlain by thick sedimentary sequences of Meso- to Neoproterozoic and Paleozoic age. The relative scarcity of Paleozoic igneous units implies that the NCC was tectonically stable throughout the early Phanerozoic [Wu et al., 2003]. From the Late Permian to Early Triassic, the NCC collided with accreted terranes of the Central Asian Orogenic Belt (CAOB). This last formational episode occurred after the Paleo-Pacific oceanic lithosphere had begun subducting beneath the northern margin of the NCC [Xiao et al., 2003].
 The SCB consists of the Archean Yangtze Craton (YC) to the northwest and the Cathaysia Block (CaB) to the southeast (Figure 1). The oldest igneous rocks found within the YC give a weighted mean U-Pb age of ∼3.2 Ga [Jiao et al., 2009] while zircons from the oldest known metamorphosed basic-ultrabasic rocks within the CaB give U-Pb ages of ∼1.85 Ga [Xiang et al., 2008]. Consensus holds that the YC and CaB were conjoined by the late Neoproterozoic. A protracted period of widespread magmatism implies that the CaB underwent multiple episodes of reactivation during the Phanerozoic. Recent in situ Hf and O isotope analyses of detrital zircons from CaB units indicate periods of crustal growth at ca. 1870 Ma, ca. 1400 Ma, ca. 1140–940 Ma, ca. 445 Ma and ca. 280 Ma, alternating with crustal recycling at ca. 1200 Ma, ca. 830 Ma and ca. 370 Ma [Li et al., 2012]. Given the multiphase history of the CaB, researchers have not been able to identify the exact boundary between the YC and the CaB.
 The formation of EC was completed by Late Jurassic time with the amalgamation of the NCC and the SCB [Zhang, 1997; Faure et al., 2001; Meng and Zhang, 2000]. Paleomagnetic data and sedimentary records indicate that convergence occurred from the Late Permian to Middle Jurassic time and resulted in a ∼70° clockwise rotation of the SCB with respect to the NCC [Zhao and Coe, 1987; Meng et al., 2005]. This collision produced the E-W-trending Qinling-Dabie-Sulu ultrahigh-pressure orogenic belt.
2.2. Tectonism Related to Pacific Plate Subduction
 EC became an active continental margin prior to the Jurassic period [Zhou and Li, 2000; Wang et al., 2006; Li and Li, 2007] due to the initiation of subduction of the Paleo-Pacific plate. Geological and geochemical evidence indicates that EC underwent multiple episodes of deformation that included high heat-flow [Hu et al., 2000], widespread tectonic extension [e.g., Ren et al., 2002] and numerous magmatic events from Late Mesozoic to Cenozoic time [e.g., Wu et al., 2005, 2011; Wang et al., 2006; Li and Li, 2007]. The subduction related deformation of EC shows distinct spatiotemporal variations. The Northeastern China Block and the NCC include wide, uninterrupted basins and multiphase magmatic events whereas the SCB includes Basin and Range-like extensional features and a relatively continuous magmatic history. The three blocks can be differentiated according to their distinct spatiotemporal units described below.
 (1) The Northeastern China Block includes a Paleozoic orogenic collage that comprises the eastern flank of the Central Asian Orogenic Belt. Mesozoic magmatic events show eastward migration trends from southeastern Mongolia to southwestern Japan [Wang et al., 2006]. Recent geochronologic data collected from Phanerozoic granitoids suggest that Northeastern China was significantly affected by subduction of the Paleo-Pacific plate during Mesozoic to Cenozoic time [Wu et al., 2011].
 (3) In contrast to the relatively stable YC, the CaB has experienced extensive tectonism [e.g., Li, 2000; Wang et al., 2005]. The CaB includes a Basin and Range–like extensional province that contrasts wider basins (widths >600 km) found in the eastern NCC. Magmatic activity affected a ∼1000 km area from the continental margin to the southeast edge of the YC [Li and Li, 2007] and temporally coincided with extensional deformation of the CaB. Geochemical analyses of 280 Ma detrital zircons from YC units reveal significant negative correlations between ϵHf(T) and δ18O values, indicating that YC igneous units formed from reworked, ancient supracrustal materials and mantle-derived magmas in an active continental margin setting [Li et al., 2012]. Equilibrium temperatures estimated by two-pyroxene geothermometry vary from 994°C to 1081°C, indicating that the Mesozoic lithospheric mantle had a relatively steep geothermal gradient [Liu et al., 2012].
3. Data and Method
3.1. Data Processing
 This study uses seismic events collected from two separate data sets. The first data set was collected from 738 permanent stations of the CNSN evenly distributed throughout the study area (circles in Figure 1), operating from July 2007 to May 2010 [Zheng et al., 2010]. In order to enhance the resolution beneath the key tectonic boundaries, a second data set was drawn from all available temporary seismic networks deployed by the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS). The IGGCAS networks consist of 4 temporary linear sub-arrays (red triangles inFigure 1), which operated sequentially from December 2007 to May 2010. Most stations were equipped with Güralp CMG-3ESP or 3T sensors (50 Hz to 30, 60 or 120 s) and REFTEK-72A or 130 data loggers. Each array was deployed for 12–18 months.
 Events with epicentral distances ≥30° from the study area were selected (Figure 2) for tomographic processing. Relative travel times of direct P- and S-waves were calculated with respect to theoretical travel times using IASP91 model assumptions [Kennett and Engdahl, 1991]. A multichannel cross-correlation technique [Vandecar and Crosson, 1990] was used to measure differential travel times in four frequency bands: 0.02–0.1 Hz, 0.1–0.8 Hz, 0.4–0.8 Hz and 0.8–2.0 Hz. Figures S1 and S2 in the auxiliary materialpresent examples of travel-time measurements for vertical and transverse components, respectively.1To reduce the influence of unevenly distributed events on the inversion step, events with higher signal-to-noise ratio were selectively processed. To enhance the regional sampling coverage within the NCC, we also selected ∼18,000 P wave and ∼10,000 S-wave arrival events from 5 other IGGCAS sub-arrays (blue triangles inFigure 1) used by Zhao et al. . In combining the data, we found that all four frequency bands have good signal-to noise ratios for P wave arrivals. The 0.4–0.8 Hz band however provided the best signal-to-noise ratio and contributed up to 59,512 relative travel-time residuals for our inversion. A total of 37,569 S-wave arrivals were selected from the 0.02–0.1 Hz band. Other frequency bands were dominated by noise. The average cross-correlation coefficient is 0.95 for both P- and S-waves. The average standard deviations in delay times from cross-correlation calculations are 0.003 s for P waves and 0.01 s for S-waves.Figure 3shows pairs of P- and S-wave relative travel-time residuals for common stations and events.
3.2. Crustal Correction
 Introduction of crustal models provides tomographic imaging procedures with a more reliable velocity distribution for the upper mantle [e.g., Ekström and Dziewonski 1998; Allen et al., 2002; Li et al., 2006; Koulakov et al., 2009]. To avoid intermingling of crustal and upper mantle signals, we employ crustal velocity and thickness models to calculate theoretical ray travel-time corrections for heterogeneity within the crust. Previous studies of the EC have developed VS crustal models based on Rayleigh wave dispersion [Huang et al., 2003, 2009] and based on local body wave tomographic imaging [Sun et al., 2008a]. Sensitivity tests of each of these crustal models indicated no significant difference between mantle model results. We therefore only present the tomographic model generated using Sun et al.'s [2008a] crustal model based on body wave tomography.
 Receiver function studies [Ai et al., 2007; Wei et al., 2011] demonstrate that the crustal VP/VS ratio varies from 1.70 to 1.84 with a mean value of 1.76 in the SCB and a mean value of 1.77 in the NCC. The mean crustal VP/VS ratio of 1.765 was used a priori to calculate correction times for P wave arrivals. To estimate the influence of VS to VP scaling, we calculated relative error for the time correction using different VP/VS ratios. Comparison of errors with those calculated using a mean VP/VS value showed that the relative error is negligible. The P wave tomographic inversions described in the following section also demonstrate that variation in the VP/VS ratio (between 1.70 and 1.84 in practice) has only a small influence on the final images generated.
 The tomographic model space is a ∼3300 km by 3300 km area (in the east-west and north-south directions, respectively) that extends to 2500 km depth, centered at 112°E, 35°N. The image volume was parameterized within a grid-space defined by regular nodes at 52 km by 52 km by 39 km. The P- and S-wave data sets were used to construct matrices based on both theoretical ray [e.g.,Allen et al., 2002] and finite-frequency kernel formulations [Dahlen et al., 2000; Hung et al., 2000]. The damped LSQR algorithm [Paige and Saunders, 1982] was then used to invert the matrix. The damping parameters were empirically determined based on the trade-off between model norm and variance reduction of data. The models discussed in this paper were constructed using damping parameters that yield a variance reduction of ∼83% for both VP and VS. Figure S3 in the auxiliary material illustrates the tradeoff between model norm and variance reduction for the VP and VSmodels based on finite frequency kernel method. To account for any baseline shift between the relative travel-time sets for different events, event corrections were included in the inversion [Allen et al., 2002].
4. VP, VS and VP/VS Ratio Models
 Tomographic results of VP, Vs and VP/Vs ratio anomalies are shown in Figures 45to 6in both horizontal map views and cross-sectional depth slices. Our results show that structures imaged using finite frequency sensitivity kernels (Figure 4) and ray-based methods are very similar except that the kernel-based models yield higher root-mean square amplitudes for P- and S-wave velocity perturbation, as expected [e.g.,Hung et al., 2004]. For convenience, we hereafter focus our discussion on the kernel-based results.
 The new P- and S-wave models are broadly consistent despite differences in the wavelengths of phases used and in station density. The VP/VSratio model negatively correlates to the P- and S-wave models but shows more small-scale variation. P- and S-wave models also show spatial patterns consistent with those identified in previous studies [e.g.,Huang and Zhao, 2006; Li et al., 2006, 2008; Sun et al., 2008a, 2008b; Tian et al., 2009; Zhao et al., 2009; Li and van der Hilst, 2010; Feng et al., 2010]. Our procedures however provide greater resolution and permit detection of certain features in more detail. Resolution tests given in section 5 demonstrate that the tomographic images provide reliable depictions of velocity heterogeneities in the upper mantle to depths of 700 km.
 An overview of the results (Figures 4 to 6) reveals strong multiscale heterogeneities (H1, H2, L1, L2, L3 in Figure 5) extending from the uppermost mantle to the mantle transition zone beneath EC. Given better data coverage for the NCC and the SCB, we focus on the structures beneath these two regions, and not Northeastern China. The VP, VS and VP/VSimages consistently display a pronounced contrast in the wavelength of the velocity anomalies occurring beneath the NCC and the SCB. In the NCC, complex, small-scale low-velocity structures are widespread in central and eastern areas. In the SCB, the YC subsurface is high velocity while the CaB subsurface is predominantly low velocity.
4.1. The North China Craton (NCC)
 The strong contrast in the between the Ordos Craton and the eastern NCC is an obvious feature of NCC tomographic images (Figures 4a, 4b, 4e, 4f, 5a, 5b, 5e, and 5f). A high-velocity region extends to >330 km depth beneath the Ordos craton (Or, H1 inFigure 5). A narrow, north-south trending, low-velocity region (with dimension of ∼800 km north-south by 200–300 km east-west) is located at the base of the central NCC lithosphere (L1 inFigure 5) [Zhao et al., 2009]. This region broadens to include much of the area beneath eastern NCC in deeper zones of the upper mantle. Cross-sectional depth slices of this low-velocity anomaly (L1;Figures 5a, 5b, 5e, and 5f) clearly illustrate that it extends from 100 km depth to the mantle transition zone beneath the central and eastern NCC. These findings are consistent with receiver function studies that imaged a similarly attenuated mantle transition zone beneath the NCC [Zhu and Zheng, 2009; Xu et al., 2011].
 A high velocity anomaly appears at depths of 430 km and 580 km, immediately east of the area beneath the Tanlu Fault Zone (TLF in Figure 4). We interpret this anomaly as the stagnant slab front of the subducted Pacific Plate [Huang and Zhao, 2006; Li and van der Hilst, 2010].
4.2. The South China Block
 In the SCB, the YC is distinguished from surrounding areas by a laterally continuous upper mantle anomaly defined by high velocity and low VP/VSratio, and extending from 100 km depth to the mantle transition zone. The largest low-velocity and high VP/VSratio anomalies are found beneath the CaB. This latter anomalous body trends SW-NE and extends from the base of the Cathaysia lithosphere (generally <100 km depth) down to the mantle transition zone. The transitional area between the YC and the CaB roughly coincides with the boundary separating Proterozoic CaB units from older YC units.
5. Resolution Tests
 The data set used to construct the tomographic model presented here is sufficient to constrain structures with dimensions of 200 km or more. Multiple resolution tests were used to verify the robustness of the solution. A synthetic 3-D velocity model generated synthetic travel-time delays for the same paths traversed by the P- and S-waves used to construct the original data set. The velocity model was then inverted to determine if the synthetic velocity structure could be recovered. For each experiment, Gaussian white noise was introduced to the synthetic relative travel-time signals to simulate noise in the data. The artificial noise was scaled to about 15% of the signal, in proportions that exceeded the standard deviation of the travel-time measurements. The same damping parameters were used in the inversions of both the synthetic and empirical data.
 We initially conducted checkerboard experiments with alternating high and low velocity anomalies of ±3% δlnVP and ±5% δlnVSwithin individual blocks measuring 208 km by 208 km in area, by 156 km in depth. The depth slices and cross-sections for VP and VS from the recovered models are shown in Figures 7 and 8. The corresponding VP/VS ratio anomalies are between ±2.1%. Depth slices for VP/VS ratio are shown in Figure 9. The tests show that the resolution is good down to 700 km depth for anomalies ≥200 km, even for data with 15% noise, given adequate regional station coverage. Inspection of the recovered images showed that the location and shape of the structures are clearly rendered. The recovered velocity anomaly amplitudes were slightly reduced.
 The second series of tests were performed to test how well the data set resolved vertical and dipping velocity anomalies. Several high- or low-velocity anomalies of varying diameter were assigned to different depth ranges (Figures 10 and 11). Velocity perturbations of ±2% δlnVP and ±3% δlnVSwere introduced to each anomaly. Inspection of the recovered images showed that the location and shape of individual anomaly structures were clearly rendered and the vertical smearing length is negligible. The vertical smearing effect was obvious however for a model with a 400 km-wide high-velocity anomaly volume (Figures 10d and 10h; labeled as H2) in the upper mantle located above a high-velocity anomaly (H3 inFigure 10) in the mantle transition zone. We interpret the high-velocity anomaly extending from the base of the lithosphere to the mantle transition zone (H2 inFigures 5c and 5g) beneath the YC as a possible smearing effect. Adjusting the dimension of H3, we also tested the influence of the minimum depth of the H2 anomaly on vertical smearing. The vertical smearing effect is weakened if the base of H2 is shallower than 200 km. This implies that high-velocity anomalies may coexist both in the upper mantle and the mantle transition zone beneath the YC.
 Seismic heterogeneities inferred from the tomographic models constructed here are strongly correlated with the major tectonic and magmatic features of EC as well as regional geodynamic events. This section discusses the salient features of the VP, VS and VP/VS models, and enlists tectonic and magmatic evidence to develop unique geodynamic interpretations for different areas of EC.
6.1. Physical Causes of the Velocity Anomalies
 Interpretations of velocity anomalies assume that spherically symmetric seismic models of the mantle reflect typical olivine phase transitions in an adiabatic P-T regime [Cammarano et al., 2003]. Mantle material can deviate from this conventional structure due to a combination of physical factors, including temperature anomalies, partial melting, bulk and volatile compositional heterogeneities, and other forms of anisotropy [Schmandt and Humphreys, 2010]. Theoretical models have shown that in the upper mantle, the scale of velocity anomalies reported by recent tomographic studies [e.g., Schmandt and Humphreys, 2010] can be primarily attributed to temperature effects and partial melting [Hammond and Humphreys, 2000; Cammarano et al., 2003]. It is predicted that a P wave velocity anomaly of ±1% corresponds to a temperature perturbation exceeding 100–200 K under dry conditions. The relative sensitivity of S- and P wave velocities to variation in mantle material is apparent as deviation in the VP/VS ratio (relative to average values). Calculation of ∂lnVS/∂lnVP gave expected values for thermal variations in the upper mantle ranging from 1.3 to 2.2 at low and high temperatures, respectively [Cammarano et al., 2003]. Partial melting in the upper mantle is predicted to attenuate velocities and cause higher ∂lnVS/∂lnVP values of ∼2.2 to 2.3 [Hammond and Humphreys, 2000; Schmandt and Humphreys, 2010].
 High VP and VS volumes, along with low VP/VS volumes (H1, H2 in Figure 5) are apparent down to depths of >300 km beneath the Precambrian Ordos block and YC. Figures 12b and 12e show an average ∂lnVS/∂lnVP value of ∼1.5 beneath these regions. These observations indicate a subsurface structure having thick lithospheric roots, lower temperatures, and a relatively depleted geochemical composition. The mantle transition zone beneath the YC exhibits high VP, VS and low VP/VS volumes (Figures 4c, 4d, 4g, 4h, 5c, 5d, 5g, and 5h). This subsurface area also becomes narrower from north to south. We interpret these features as evidence of the stagnant slab of the Paleo-Pacific Plate subducted beneath Southeast China during the Mesozoic [e.g., Li and van der Hilst, 2010].
 The central and eastern NCC and CaB exhibit low VP and VS volumes, and high VP/VS volumes (Figures 4 and 6), which are typical of young igneous provinces. The large low-velocity anomaly observed in this region indicates warm mantle material associated with large-scale Mesozoic to Cenozoic igneous features. A majority of the Cenozoic rift systems and basalt deposits in the NCC, including the Shanxi-Shanxi Rift System and the Hannuoba volcanics [Zhao et al., 2009] are located above the low-velocity zones imaged here. The horizontal outline of the low-velocity zone beneath the SCB also coincides with an extensive igneous province [e.g.,Li and Li, 2007]. Figures 12c and 12d and Figures 12f and 12g show that ∂lnVS/∂lnVP values are >2.2 for these magmatic provinces, indicating that partial melting contributes to velocity anomalies beneath these regions.
 In general, we interpret the observed VP, VS and VP/VS variation as evidence of laterally varying temperature differences due to variable lithospheric thickness, asthenospheric features and stagnant slab volumes in the mantle transition zone. Areas having extremely low VP and VS anomalies and high VP/VS ratio anomalies reflect partial melting beneath EC [e.g., Goes and van der Lee, 2002; Schmandt and Humphreys, 2010].
6.2. Potential Causes for Spatial Variations in EC Evolution
 Geological, geophysical and geochemical evidence suggest that the geodynamic evolution of EC is primarily a controlled by Mesozoic to Cenozoic subduction of the Paleo-Pacific plate [Wu et al., 2005; Wang et al., 2006; Huang and Zhao, 2006; Li and Li, 2007; Sun et al., 2007; Zhao et al., 2007; Lin et al., 2008]. Differences in the scale of anomalous volumes imaged beneath eastern NCC and the SCB imply distinct geodynamic processes for these regions. Subduction structures and processes appear to be laterally continuous beneath the CaB, but fragmented beneath the eastern NCC. For a given subduction zone, spatial variation in the continental plate's evolution depends on the water content and geometry of the subducting plate, as well as the intrinsic properties of the continental block. Figure 13 presents a hypothetical series of events that explain the causes for spatial variability in EC's response to different stages of subduction.
 Differing geometries of the fragmented subducting plate may cause spatial variations in EC evolution and lateral variation in the water-content of the subducting slab would then strengthen these heterogeneities. While the lithospheric thickness of the CaB prior to the Mesozoic is unknown, the NCC and the YC persisted as stable cratons with thick lithospheric roots until the early Paleozoic [e.g.,Menzies et al., 1993; Griffin et al., 1998]. With the subduction of the Pacific Plate and eastward retreat of the slab [e.g., Li and van der Hilst, 2010], subduction-induced mantle convection beneath EC would induce a temperature increase and partial melting of the uppermost mantle, resulting in the reactivation of EC lithosphere. Previous tomographic models [e.g.,Huang and Zhao, 2006; Li and van der Hilst, 2010] and regional receiver-function images of the mantle transition zone [e.g.,Ai and Zheng, 2003; Zhu and Zheng, 2009; Chen and Ai, 2009; Xu et al., 2011] indicate that the slab front of the Pacific Plate extends well into the area beneath the SCB (∼105°E), but only reaches the eastern boundary of the NCC (∼118°E). The impinging slab front would therefore introduce stronger subcrustal material (the slab itself) and enhance convergent flow and thermal convection beneath the SCB, but not beneath the eastern NCC. The offset in the slab front between SCB and NCC explains the larger, laterally continuous low velocity and VS/VPratio anomalies observed in the CaB upper mantle, but not beneath the NCC. The Basin and Range-style rift systems of the SCB and the wide, uninterrupted basins of the NCC indicate that a heterogeneous compressional stress-field affects the former area, but not the latter.
 The outlines of velocity anomalies are spatially consistent with major tectonic units and boundaries in the SCB and the CaB. The correspondence of geophysical and geological features suggests the intrinsic properties of both the NCC and SCB also influence mantle-flow patterns. The NCC includes a thick Archean lithospheric component that was attenuated and rifted from Mesozoic to Cenozoic time [e.g.,Griffin et al., 1998; Chen et al., 2008; Zheng et al., 2008; Zhu and Zheng, 2009]. U-Pb zircon geochronology dates the CaB as Neoproterozoic in age, indicating that it is younger than the NCC [Yu et al., 2006]. Recent geodynamic models demonstrate that high viscosity conditions or a cold thermal state in the mantle may create localized weak zones when a craton is subjected to extensional stress [Lu et al., 2011]. Given these theoretical constraints, subduction of the Pacific Plate would generate smaller-scale anomalies beneath the older NCC, but produce larger, laterally continuous anomalies beneath the younger CaB.
 We present new 3-D tomographic models of VP, VS and VP/VS ratio anomalies that demonstrate strong, multiscale heterogeneities in the upper mantle beneath EC. Resolution tests for different depths and the prominent velocity anomalies identified by our model verify that the tomographic images can resolve velocity heterogeneities in the upper mantle to depths of 700 km.
 The anomalous volumes observed beneath the NCC and SCC exhibit a pronounced contrast in scale. These features demonstrate that different areas of EC have differing histories associated with Pacific Plate subduction. Seismic heterogeneities evident from tomographic models are found beneath major tectonic and magmatic features of EC, providing a link between major surface features and larger-scale geodynamic events. In general, we suggest that most of the observed VP, VS and VP/VS ratio anomalies reflect variations in lithospheric thickness, asthenospheric conditions, stagnant slab volumes and the associated temperature perturbations. The strongest observed anomalies are likely due to areas of partial melting beneath EC.
 We attribute the spatial variation in EC's response to ongoing subduction related deformation to two mechanisms. First, the subducting slab reaches further beneath the SCB in the westward direction and thus subjected this region to greater tectonic stress, introduced more subduction related material and strengthened thermal convection. The slab does not significantly impinge past the eastern boundary of the NCC, resulting in fewer observed anomalies. Second, the eastern NCC consists of an older Archean lithospheric component having a colder thermal state, stronger rheology and smaller-scale heterogeneities. In contrast, the younger Neoproterozoic CaB has a warmer thermal state and weaker rheology. This latter region therefore exhibits more continuous deformational features relative to the NCC.
 We thank Xiufen Zheng, Mingming Jiang, Yinshuang Ai and Ling Chen for their help with data collection. Discussions with Mathias Obrebski, Ling Chen, Wei Lin, and Gang Lu improved the manuscript. We thank Editor Thorsten Becker, the associate editor, Yang Shen and an anonymous reviewer for very helpful reviews. Waveform data are provided by Seismological Laboratory, IGGCAS and Data Management Centre of China National Seismic Network at Institute of Geophysics, China Earthquake Administration. This research was financially supported by Chinese Academy of Sciences (grant KZCX2-YW-QN102), the NSFC grant 90914011 and 40974030.