Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
1 Ambient seismic noise tomography of the southern East Sea (Japan Sea) and the Korea 2 Strait 3 : Group velocity maps for the southern East Sea and the Korea Strait with three distinct velocity 4 anomalies 5 6 Sang-Jun Lee1,a, Junkee Rhie1,b,*, Seongryong Kim1,2,c, Tae-Seob Kang3,d and Gi Bom Kim1,e 7 8 1 9 Korea. School of Earth and Environmental Sciences, Seoul National University, Seoul 151-742, 10 2 Research School of Earth Science, Australian National University, Canberra, ACT, Australia 11 3 Department of Earth and Environmental Sciences, Pukyong National University, Busan 608- 12 737, Korea. 13 a [email protected] , b [email protected], c seongryong,[email protected], d [email protected], 14 e [email protected] 15 16 * Corresponding author : Junkee Rhie E-mail: [email protected] Tel: +82-2-880-6731 School of Earth and Environmental Sciences Seoul National University Seoul 151-742, Korea 1 17 Abstract 18 Group velocity maps were derived for the southern East Sea (Japan Sea) and the Korea Strait 19 (Tsushima Strait) for the 5-36 s period range, which is sensitive to shear wave velocities of the 20 crust and the uppermost mantle. Images produced in our study enhance our understanding of 21 the tectonic evolution of a continental margin affected by subducting oceanic slabs and a 22 colliding continental plate. The seismic structure of the study area has not been described well 23 because seismic data for the region are scarce. In this study, we applied the ambient noise 24 tomography technique that does not rely on earthquake data. We calculated ambient noise 25 cross-correlations recorded at station pairs of dense seismic networks located in the regions 26 surrounding the study area, such as the southern Korean Peninsula and southwestern part of 27 the Japanese Islands. We then measured the group velocity dispersion curves of the 28 fundamental mode Rayleigh waves from cross-correlograms and constructed 2-D group 29 velocity maps reflecting group velocity structure from the upper crust to uppermost mantle. 30 The results show that three distinct anomalies with different characteristics exist. Anomalies 31 are located under the Ulleung Basin (UB), the boundary of the Basin, and the area between 32 Tsushima Island and the UB. 1-D velocity models were obtained by inversion of dispersion 33 curves that represent vertical variations of shear wave velocity at locations of three different 34 anomalies. The 1-D velocity models and 2-D group velocity maps of lateral variations in shear 35 wave group velocities show that the high velocity anomaly beneath the UB originates from 36 crustal thinning and mantle uplift. Confirming the exact causes of two low velocity anomalies 37 observed under the UB boundary and between Tsushima Island and the UB is difficult because 38 additional information is unavailable. However, complex fault systems, small basins formed 39 by faulting, and deep mantle flow can be possible causes of the existence of low velocity 40 anomalies in the region. 2 41 Keywords: East Sea, Ulleung Basin, Korea Strait, Ambient noise tomography, Group velocity 42 1. INTRODUCTION AND BACKGROUND 43 There have been many previous studies focusing on relatively large-scale features in the upper 44 mantle of the northeastern region of Asia (Huang and Zhao, 2006; Zheng et al., 2011; Wei et 45 al., 2012). Along with the large-scale studies, detailed research on inland regions has been 46 performed using data recorded at dense seismic networks in the Japanese Islands (JI) and 47 Korean Peninsula (KP) (Kang and Shin, 2006; Yoshizawa et al., 2010; Guo et al., 2013; Liu et 48 al., 2013). The seas between the KP and JI are of particular interest to seismologists because 49 surrounding plate interactions in the past triggered multiple tectonic events (Fig. 1., Chough 50 and Barg, 1987; Tamaki et al., 1992; Kim et al., 2008; Son et al., 2013). However, the detailed 51 seismic structures have not been understood well and only limited information from active- 52 source seismic experiments is available. Lee et al. (2001) proposed a seismic stratigraphic 53 interpretation of the Ulleung Basin (UB) by analyzing multi-channel seismic reflection profiles. 54 Lee et al. (1999) and Kim et al. (2011) revealed the crustal structure and volcanic history of 55 the UB. Cho et al. (2004), Sato et al. (2006), and Lee et al. (2011) determined the marginal 56 structure of our study region and presented possible tectonic causes. These studies yielded well- 57 resolved shallow (sediment-to-crustal) structures along profiles. But most of profiles sampled 58 the UB, while relatively few profiles sampled marginal areas, limiting our understanding of the 59 region. The maximum resolvable depth of these studies is near the depth of the Moho, further 60 limiting the applicability of existing research. 61 Passive seismic methods can help overcome these limitations. Due to a lack of seismic stations 62 on the ocean floor, methods that use body waves to reveal seismic structures are not considered 63 appropriate. Therefore, surface wave method can be used as an alternative to study seismic 64 velocity structures in a given region. Recent surface wave studies in the southern East Sea (ES) 3 65 and Korea Strait (KS) help characterize our smaller study area. Yoshizawa et al. (2010) used 66 interstation dispersion data measured from earthquake waveforms and Zheng et al. (2011) used 67 surface wave dispersion data obtained from ambient noise. Both investigations used data 68 recorded at multiple stations in Northeast Asia to generate high-resolution images beneath the 69 southern ES and KS. However, the studies utilized only one station in the KP. By using multiple 70 stations in the KP, we can enhance the resolution of the models for our study area. 71 Relatively long inter-station distances potentially restrict use of high-frequency dispersion data 72 for studying the shallow crustal structure of the southern ES and KS. Lack of permanent 73 seismic stations within the study area also potentially limit investigation of deeper structures 74 in the basin. However, the density of the seismic network surrounding the study area is very 75 high, even though azimuthal coverage is not perfect. Ambient noise tomography methods can 76 address coverage problems in this situation because station pair paths create an intersecting 77 network beneath the study area. Ambient noise tomography has been successfully applied in 78 Korea and Japan before (Kang and Shin, 2006; Guo et al., 2013). Inter-station distances in the 79 target region are a possible concern because of potential problems extracting Green’s functions. 80 Lin et al. (2006) addressed the concern by showing it is possible to extract Green’s functions 81 by cross correlating station paths beneath the ocean. Also, feasibility tests verified that 82 tomographic methods using array data surrounding oceans are effective (Pawlak et al., 2011; 83 Zheng et al., 2011). Therefore, we selected ambient noise tomography to study group velocity 84 structures and lateral seismic velocity changes in the region. 85 86 2. TECTONIC SETTING OF THE EAST SEA AND KOREA STRAIT 87 East Asia is located at the eastern margin of the Eurasia plate, in a tectonic setting complicated 88 by the subducting Pacific and Philippine Sea plates and the colliding India plate. The India– 4 89 Eurasia continental collision and the East Asian subduction have created many back-arc basins 90 and Cenozoic normal and strike-slip faults in the study region (Schellart and Lister, 2005; Wei 91 et al., 2012). Subducting plates cause complex and various tectonic activities such as 92 earthquakes in and around the ES and volcanism along the JI, and numerous marginal basins 93 are formed as a result of slab rollback processes (Schellart and Lister. 2005). Beneath the 94 Kyushu region, a low velocity anomaly exists that appears to represent upwelling from a mantle 95 wedge disturbed by a subducting slab (Seno, 1999; Sadeghi et al., 2000; Yoshizawa et al., 96 2010). The hot upwelling corresponds to volcanic chains in the Kyushu region, including the 97 Aso volcano (Zheng et al., 2011). 98 The ES is a back-arc basin that started to form in the late Oligocene (Tamaki et al., 1992). Three 99 deep basins (Japan, Yamato, and Ulleung basins) and some topographic highs (Korean plateau, 100 Yamato rise, and Sea Mountains) are the main features of the ES. The basins were formed 101 during tectonic extension, and elevated areas are remnants of continental crust or volcanic 102 deposits. After the ES opened, the KS formed by clockwise rotation of southern Japan caused 103 by northward motion of the Philippine Sea plate 15–16 Ma (Kim et al., 2008; Son et al., 2013). 104 Crustal thickness of the ES and KS varies from place to place. The crust of the ES is generally 105 thicker than normal oceanic crust (~7 km) and the thickness at UB is more than double the 106 average (~20 km). In contrast, the KS consists of relatively thick continental crust (~30km) 107 with a relatively thin sediment layer. Crustal thickness is greatest at the KP and JI (~36 km) 108 (Chough and Barg, 1987; Cho et al., 2004; Gil’manova and Prokudinf, 2009; Kim et al., 2011). 109 Many fault systems exist around the UB and KS. Along the western margin of the UB, there 110 are the Hupo and Ulleung faults and associated small folds and thrust faults. The Hupo fault 111 running along the continental slope of the eastern Korean continental margin is a long, narrow 112 normal fault (~140 km long and ~3–4 km wide) with vertical displacement up to 1 km. The 5 113 Ulleung fault is a strike-slip fault running N-S to NNE-SSW (Kang et al., 2013; Yoon et al., 114 2014). The Tsushima-Goto fault system is a major tectonic line around the Tsushima Island 115 (TI), with normal fault systems dominating the western part, and thrust faults and folds 116 dominating the eastern side. The area also contains compressional structures like the Dolgorae 117 Thrust Belt (DTB) northeast of TI, and the San`in fold belt between TI and JI (Kim et al., 2008; 118 Son et al., 2013). 119 120 3. DATA COLLECTION AND MODELING METHODS 121 Continuous waveforms recorded at 60 broadband seismic stations in 2008 were used for our 122 study. We obtained vertical component waveforms from 28 stations in the KP deployed by the 123 Korea Meteorological Administration (KMA) and the Korea Institute of Geoscience and 124 Mineral Resources (KIGAM). In addition, we used continuous waveforms from 31 stations 125 belonging to the F-net broadband seismograph network operated by the National Research 126 Institute for Earth Science and Disaster Prevention (NIED) and from 1 station of the Global 127 Seismic Network (GSN) (Fig. 1). 128 To understand group velocity structure in the study area, we constructed group velocity maps 129 for several periods (5, 10, 20, 25, 30, and 36 s) using given waveform data. First, we calculated 130 the noise cross-correlation function (NCF, which represents Green’s function of each station 131 pair) for all possible station pairs. Second, the group velocities of fundamental mode Rayleigh 132 waves were measured for different periods with a 1 s increment. Finally, theoretical travel times 133 for all station pairs were inverted for 2-D group velocity maps of specific periods. The NCF 134 representing seismic wave propagation between two stations can be extracted from the cross- 135 correlation of ambient noise recorded at two given stations (Shapiro and Campillo, 2004; 136 Shapiro et al., 2005; Bensen et al., 2007). This method is only valid when the ambient noise 6 137 consists of diffusive wave fields. Specifically, the temporal and spatial distribution of the noise 138 sources should be random. Therefore, reducing effects of transient but coherent signals (e.g., 139 earthquakes) is important, and two methods are widely used. The first method normalizes 140 amplitudes of the waveforms in the time domain. We utilized the second method, which 141 removes the portion of waveforms likely contaminated by earthquakes or unknown localized 142 sources. 143 NCFs were extracted using the following procedures: (1) Continuous vertical component 144 waveforms were divided into 1 h windows with 30 min overlap. (2) The mean, trend, and 145 instrument response of each window were removed. (3) Sampling rates of waveforms were 146 decreased to 1 sample per second (sps) if the sampling rate of the original waveform was not 1 147 sps. (4) A band pass filter between 0.01 and 0.45 Hz was applied. (5) The windows that 148 appeared to be contaminated by earthquakes were removed. We discarded windows if their 149 peak amplitudes were 10 times larger than the median root mean square (RMS) of all windows 150 for each station. (6) Cross-correlation was calculated for all possible pairs of 1 h windows. In 151 this step, spectral whitening was applied by dividing a 40-point moving average in the 152 frequency domain. (7) The final correlogram for each station pair was calculated by averaging 153 all remaining correlograms that satisfied the condition in step (5). (8) The NCF was extracted 154 from the correlogram. The correlogram consists of causal (positive lag) and acausal (negative 155 lag) sections. The two sections represent the NCFs from one station to another. When ambient 156 noise is completely diffusive, both sections appear identical. In many cases, however, the 157 distribution of noise sources is not completely random. Therefore, we must pick the optimal 158 NCF for a given station pair between the causal section, acausal section, or average of two 159 sections. Strong and continuous seismic sources are known within the study area, and they 160 include the well-documented Kyushu microseisms that are explained by activity of the Aso 7 161 volcano. The detailed procedures for reducing the effects of the Kyushu microseisms are 162 described in a previous study (Zheng et al., 2011). We used the same procedures, but took the 163 location of the Aso volcano as the source location of the Kyushu microseisms. We manually 164 checked the correlograms and only included NCFs that were less contaminated by the Kyushu 165 microseism activity. 166 To measure group velocities from the NCF, the multiple filter technique of Herrmann and 167 Ammon (2002) was applied. We filtered the NCF using a narrow Gaussian filter with a specific 168 center period and calculated an envelope. By repeating this process for different periods we 169 constructed Frequency Time Analysis (FTAN) diagrams, which present energy distribution as 170 a function of time and period (Fig. 2). The group velocity dispersion curve of the fundamental 171 mode Rayleigh wave can be obtained from FTAN diagrams by picking corresponding peaks at 172 each period. The group velocity was measured only if the inter-station distance of the station 173 pair was 3 times longer than the wavelength at the given period. To reduce possible errors, we 174 manually measured the group velocities when the corresponding peaks were clearly identified 175 in the diagram. The total number of group velocity measurements varied with the periods, but 176 more than 1000 velocities were measured in all cases. The average group velocity dispersion 177 curve of the region was calculated by bootstrapping, and the average values were used as 178 reference values for 2-D group velocity maps. 179 A nonlinear 2-D tomographic method was applied for the inversion. The propagation paths 180 between the stations were updated by using the fast marching method (FMM) (Sethian and 181 Popovici, 1999; Rawlinson and Sambridge, 2005; Rawlinson, 2005). The subspace method 182 (Kennett et al., 1988) was used to find optimal model perturbations in each iteration. The 183 combination of FMM and the subspace method has been verified to provide stable results in 184 many previous studies (Rawlinson and Sambridge, 2005; Saygin and Kennett, 2010; Kim et 8 185 al., 2012). We selected optimal parameters for damping and smoothing by choosing an 186 inflection point in the trade-off curve between the RMS misfit and the model variance for each 187 period. The final damping parameter varies from 10 to 20 for different periods and the 188 smoothing parameter is fixed to be 10 for all periods. 189 190 4. RESULTS AND DISCUSSION 191 4.1. Reliability and verification 192 The group velocity maps reflecting structure beneath the study region have been developed for 193 periods at 5, 10, 20, 25, 30, and 36 s (Fig 3). To evaluate the reliability of the maps, we first 194 conducted checkerboard tests. We constructed testing models with three different checker sizes 195 (0.6°x0.6°, 1°x1° and 1.3°x1.3°) with ±10% velocity perturbations. Synthetic arrival time data 196 were produced for the same distributions of inter-station paths with observations. The synthetic 197 data were inverted for the group velocity maps using the same inversion scheme and the 198 resulting images were compared with the original testing models (Fig 4). The test results show 199 that checkers with even the smallest size are well recovered in the KP and JI where the ray 200 density is high and azimuthal coverage is good (Fig 4). In the ES and the KS, structures with 201 sizes greater than 1°x1° are resolvable. However, the recovered images using smaller checkers 202 are smeared in the NW-SE direction around the ES and the KS. The pattern implies that small- 203 scale anomalies shown in the resulting group velocity maps could be artifacts or may be 204 extended in the NW-SE direction. 205 To verify the reliability of our results, the group velocity maps were compared with the 206 previous results for the southern KP and the southwestern JI. A significant low velocity 207 anomaly is observed in the Kyushu region of JI in this study, consistent with an ambient noise 208 phase velocity map constructed from a previous study (Guo et al., 2013). The low velocity 9 209 anomaly at Kyushu coincides with volcanic regions at the surface. In contrast, group velocity 210 structures obtained in this study around the Shikoku and the Nankai regions look quite different 211 from the results of Guo et al. (2013). The discrepancy may relate to poor resolution of our map 212 in these regions, as shown in our checkerboard tests. For the KP, the absolute level of velocity 213 perturbation is less than 4% for all periods, indicating that lateral variation of seismic structures 214 beneath the KP are not significant. Previous results of ambient noise tomography in the KP 215 highlight shallow crustal structure (Kang and Shin, 2006; Cho et al., 2007, Choi et al., 2009) 216 and show that velocities at the northwestern and the southeastern parts of the KP are lower than 217 other places. A similar pattern is observed in our group velocity map for the 5 s period (Fig 3a). 218 219 4.2. Regional velocity anomalies 220 The group velocity maps of different periods show three distinctive anomalies beneath the 221 southern ES and KS (Fig 3). Anomaly 1 is a significant high velocity zone beneath the UB for 222 periods between 20 and 30 s. Anomaly 2 is a low velocity region observed at 5 and 10 s periods 223 in continental shelves east of the KP and in the northwest of the southwestern JI, surrounding 224 the UB. Anomaly 3 is the most prominent low velocity anomaly around the KS, and clearly 225 appears at much wider period ranges. The three anomalies are identified based on lateral 226 variation of group velocities for different periods. A 1-D shear wave velocity model was 227 calculated from our measured average dispersion curve. The model shows Rayleigh wave 228 group velocities for 5 and 36 s periods are highly sensitive to shear wave velocities at depths 229 of 5 and 40 km, respectively (Fig. 5). Applying this information, the calculated depths of 230 Anomalies 1, 2, and 3 are roughly 18–30, 5–8, and 5–40 km, respectively. 231 In order to estimate more exact variations of shear wave velocities with depth, we performed 232 1-D depth inversions using local dispersion curves of group velocity for 4 different locations. 10 233 We selected locations to characterize and compare regions of velocity anomalies shown in Fig. 234 3: the basement of the UB (Fig. 6a) for Anomaly 1, continental shelves around JI (Fig. 6b) for 235 Anomaly 2, the center of the low-velocity anomaly in the KS (Fig. 6c) for Anomaly 3, and the 236 southern sea of the KP (Fig. 6d) as a reference. The local dispersion curves were created by 237 spline interpolation using the group velocity maps in the range of 5 to 36 s periods with 1 s 238 increments (Fig. 3). The iterative inversion method of Julia et al. (2000) was utilized to obtain 239 shear velocity models. For the starting model of each inversion, we used slightly modified 240 models from CRUST1.0 to allow good convergence. After several trials using smoothing 241 values in the 0-10 range, we constrained the smoothness to 1.0 for the inversions. Synthetic 242 dispersion curves fit the observed local dispersion curves with root-mean-square errors less 243 than 0.025 km/s (Fig 6). We produced 4 different 1-D models representing depth variations of 244 shear wave velocity at 4 locations and call them Model 1 through Model 4. 245 Possible causes of each velocity anomaly are varied. For Anomaly 1, the lateral extent of the 246 anomaly is confined within the UB. The related Model 1 has greater velocity than reference 247 Model 4 over all depth ranges. Because surface wave dispersion is smoothly sensitive to shear 248 velocities across depths (Fig. 5), the increased velocities and rapid velocity changes with depth 249 likely indicate a shallower crust/mantle transition. Previous observations of thinner crust in the 250 region support this explanation and the thinner crust is explained by extension following the 251 opening of the ES (Kim et al., 1998; Sato et al., 2006; Park et al 2009; Lee et al., 2011). We 252 infer from our model and supporting study findings (Zheng et al., 2011) that Anomaly 1 results 253 from locally uplifted mantle structure. At the location of Anomaly 1, the effect of the water 254 layer may not be negligible because the depth of water is around 2 km (e.g., Huang et al., 2014). 255 To check the reliability of the features resolved by the inverted model for this location, we 256 performed an inversion with a water layer as well (Herrmann and Ammon, 2002). For this 11 257 purpose, the previously inverted model was taken as the initial model. Since modeling the 258 dispersion curve at short periods (5-7 s) gives numerically unstable results, data for periods 259 longer than 7 s were used. The results showed that the inverted model with the water layer was 260 comparable with the previous model, except for the shallow depth (~15 km) (Fig 6a). However, 261 the increased velocity in the upper part of the newly inverted model provided additional 262 evidence in support of the described features, in spite of the introduced water layer modifying 263 the model. For other anomalies, the effect of the water layer was negligible because the 264 thickness of the water layer was less than a couple of hundred meters. 265 For Anomaly 2, the related Model 2 shows relatively lower velocity than reference Model 4 at 266 depths shallower than 10 km. In map view, Anomaly 2 surrounds the UB and appears only in 267 the ocean. Anomaly 2 may represent distribution of sediments, but the sediment layers along 268 the eastern Korean continental margin are thinner than sediments within the UB (Chough and 269 Barg, 1987; Sato et al., 2006; Gil’manova and Prokudin, 2009; Park et al 2009; Lee et al., 2011). 270 Other possible explanations for low velocity anomalies at shallow crust of the UB boundary 271 are faults and related structures, such as the Ulleung, Hupo, and Yangsan faults along the 272 eastern coast of the southern KP and western margin of the UB. Faults and folds are also 273 distributed between Tsushima and Oki islands (Kim et al., 2008; Gil’manova and Prokudin, 274 2009; Yoon et al., 2014). 275 In map view, Anomaly 3 appears inside Anomaly 2 and the velocity of Model 3 is slower than 276 that of reference Model 4 in the crust and upper mantle. Anomaly 3 and Anomaly 2 may share 277 the same origin at shallow depths, but Anomaly 3 could extend deeper into the upper mantle. 278 Therefore, similar to Anomaly 2, sedimentary structures are not the likely cause of the anomaly. 279 Significant fold and fault systems near Anomaly 3, such as the Tsushima-Goto fault, Dolgorae 280 Thrust Belt, and San’in folded zone (Kim et al., 2008; Son et al., 2013; Yoon et al., 2014), 12 281 could be an origin of the low velocity anomaly in the shallow crust. However, the fold and fault 282 systems do not likely explain the low velocity anomaly extending to the upper mantle. A 283 complementary component that may explain the deep low velocity anomaly is mantle 284 upwelling flow. A previous study reported that low velocity anomalies exist in the upper mantle 285 beneath the eastern and western boundaries of the ES and they indicate upwelling flow, from 286 slab dehydration at several depths, possibly caused by phase transitions and stagnation of the 287 slab (Zhao et al., 2007; Zheng et al., 2011). The surface trace of the upwelling flow reported 288 by Zheng et al. (2011) closely matches the low velocity region within our study area. This 289 mantle upwelling may facilitate hydrothermal fracturing. From this perspective, Anomaly 3 290 can be explained by significant fold and fault systems that are the local center of enhanced 291 hydrothermal activities originating from upwelling mantle flow. The lack of an obvious low 292 velocity anomaly in the upper mantle corresponding to Anomaly 2 may be because of low 293 resolution of long period surface waves that sample the upper mantle, or because of relatively 294 weak hydrothermal activities in the given area. 295 The relationship between the group velocity structure and seismicity supports our low velocity 296 anomaly hypothesis. We related the group velocity map at 5 s and seismicity from the ISC- 297 GEM Catalog (Storchak et al., 2013) (Fig. 7). Compared to the UB and the surround marginal 298 area, seismicity is clearly more focused in the low velocity region. Generally, shear wave 299 velocity is negatively correlated to the thermal structure and amount of melt in the crust. In 300 contrast, seismicity is lower in thermally enhanced regions (Kim et al., 2012). However, high 301 seismicity occurs in the low velocity region and low seismicity occurs in the high velocity 302 region of our study area, implying that the seismicity pattern in our study area cannot be 303 explained by thermal structure only. Even if the pre-existing fault systems are more developed 304 in the KS region (Son et al., 2013; Yoon et al., 2014), they simply do not explain the high 13 305 seismicity in the predominately low velocity region. A suggested cause of weak crustal material 306 and increased seismicity could be water that is continuously supplied to the crust by mantle 307 upwelling flow (Hasegawa et al., 2005). We speculate that water supplied by upper mantle 308 upwelling enhances seismic activity while reducing seismic velocity by increasing melts and 309 heat. 310 311 5. CONCLUSION 312 To study seismic group velocity structure of the southern ES and KS, 2-D group velocity maps 313 and 1-D shear wave velocity models were calculated by using ambient noise data. Our 314 observations show that a significant high velocity anomaly exists in the UB and low velocity 315 anomalies of different periods exist in marginal areas. The high velocity anomaly can be 316 explained by locally uplifted mantle structure beneath the UB. We can conclude that the crustal 317 thickness of the UB is relatively thin compared to surrounding regions. In contrast, the shallow 318 crust of the UB boundary has a low Rayleigh wave group velocity and a localized low velocity 319 anomaly extends from the crust to uppermost mantle in the region between Tsushima and the 320 UB. We suspect that the low velocity anomalies described in our study might be associated 321 with hydrothermal activities of mature fold and fault systems in the area. Comparing our group 322 velocity map with regional seismicity, we demonstrate that zones with high seismicity usually 323 have low velocities. This observation may be explained by weakened crustal strength combined 324 with a reduced seismic velocity caused by water supplied from the upwelling mantle. 325 326 Acknowledgement 14 327 We thank the operators of KMA, KIGAM and NIED for making their data publicly available. 328 This work was funded by the Korean Meteorological Administration and Development 329 Program under Grant CATER 2012-5051. 330 331 15 332 References 333 Bensen, G.D., Ritzwoller, M.H., Barmin, M.P., Levshin, A.L., Lin, F., Moschetti, M.P., Shapiro, 334 N.M., and Yang, Y., 2007, Processing seismic ambient noise data to obtain reliable broad- 335 band surface wave dispersion measurements. Geophysical Journal International, 169, 1239– 336 1260. 337 338 Bourova, E., Yoshizawa, K., and Yomogida, K., 2010, Upper mantle structure of marginal seas 339 and subduction zones in northeastern Eurasia from Rayleigh wave tomography. Physics of 340 the Earth and Planetary Interiors, 183, 20–32. 341 342 Cho, H., Kim, H., Jou, H., Hong, J., and Baag, C., 2004, Transition from rifted continental to 343 oceanic crust at the southeastern Korean margin in the East Sea (Japan Sea). Geophysical 344 Research Letters, 31, L07606. 345 346 Cho K.H., Herrmann, R.B., Ammon, C.J., and Lee, K., 2007, Imaging the Upper Crust of the 347 Korean Peninsula by Surface-Wave Tomography. Bulletin of the Seismological Society of 348 America, 97, 198-207. 349 350 Choi, J., Kang, T., and Baag, C., 2009, Three-dimensional surface wave tomography for the 351 upper crustal velocity structure of southern Korea using seismic noise correlations. 352 Geosciences Journal, 13, 423 – 432. 353 354 355 Chough, S.K. and Barg, E., 1987, Tectonic history of Ulleung basin margin, East Sea (Sea of Japan). Geology, 15, 45-48. 16 356 357 358 Gil’manova, G. Z. and Prokudin, V. G., 2009, Density Model of the Lithosphere in the Tsushima Basin, Sea of Japan. Russian Journal of Pacific Geology, 2009, 3, 596–602 359 360 Guo Z., Gao, X., Shi, H., and Wang, W., 2013, Crustal and uppermost mantle S-wave velocity 361 structure beneath the Japanese islands from seismic ambient noise tomography. 362 Geophysical Journal International. 193, 394–406 363 364 Hasegawa, A., Uchida, N., Igarashi, T., Matsuzawa, T., Okada, T., Miura, S., and Suwa, Y., 365 2005, Asperities and quasi-static slip on the subducting plate boundary east off Tohoku, NE 366 Japan. In: Dixon, T. and Moore, C. (eds), The Seismogenic Zone of Subduction Thrust 367 Faults. Columbia University Press, New York. 368 369 370 Herrmann, R.B. and Ammon, C.J., 2002, Computer programs in seismology: Surface waves, receiver functions and crustal structure. Version 3.30, Saint Louis University. 371 372 373 Huang, J. and Zhao, D., 2006, High-resolution mantle tomography of China and surrounding regions. Journal of Geophysical Research, 111, B09305. 374 Huang, Y. C., Yao, H., Wu, F. T., Liang, W. T., Huang, B. S., Lin, C. H., & Wen, K. L., 2014, 375 Crustal and upper mantle S-wave velocity structures across the Taiwan Strait from ambient 376 seismic noise and teleseismic Rayleigh wave analyses. Journal of Asian Earth Sciences, 81, 377 38-52. 378 379 Huang, Z., Zhao, D., Hasegawa, A., Umino, N., Park, J., and Kang, I., 2013, Aseismic deep 17 380 subduction of the Philippine Sea plate and slab window. Journal of Asian Earth Sciences, 381 75, 82–94 382 383 Julia, J., Ammon, C. J., Herrmann, R. B., and Correig, A. M., 2000, Joint inversion of receiver 384 function 385 International, 143(1), 99-112. and surface wave dispersion observations.Geophysical Journal 386 387 Kang T. and Shin, J., 2006, Surface-wave tomography from ambient seismic noise of 388 accelerograph networks in southern Korea, Geophysical Research Letters. 33, L17303. 389 390 Kang, Y., Kim, W., and Kang, T., 2013, Relocation of earthquakes beneath the East Sea of 391 Korea: uncertainty of hypocentral parameters caused by refracted waves. Geosciences 392 Journal, 17, 173 – 182. 393 394 395 Kennett B. L. N., Sambridge, M. S., and Williamson, P. R., 1988, Subspace methods for large inverse problems with multiple parameter classes. Geophysical Journal, 94, 237-247. 396 397 Kim, G.B., Yoon, S.H., Chough, S.K., Kwon, Y.K., and Ryu, B.J., 2011,Seismic reflection 398 study of acoustic basement in the South Korea Plateau, the Ulleung Interplain Gap, and the 399 northern Ulleung Basin: Volcano-tectonic implications for Tertiary back-arc evolution in 400 the southern East Sea. Tectonophysics, 504, 43. 401 402 Kim, H., Song, C., Kim, J., Son, M., and Kim, I., 2008, Tertiary Geological Structures and 403 Deformation History of the Southern Tsushima Island, Japan. Journal of the Geological 18 404 Society of Korea. 44, 175-198. 405 406 Kim S., Nyblade, A.A., Rhie, J., Baag, C., and Kang, T., 2012, Crustal S-wave velocity 407 structure of the Main Ethiopian Rift from ambient noise tomography, Geophysical Journal 408 International. 191, 865–878. 409 410 Lee, G.H., Kim, H.J., Suh, M.C., and Hong, J.K., 1999, Crustal structure, volcanism, and 411 opening mode of the Ulleung Basin, East Sea (Sea of Japan). Tectonophysics, 308, 503– 412 525. 413 414 Lee, G.H., Kim, H.J., Han, S.J., and Kim, D.C., 2001, Seismic stratigraphy of the deep Ulleung 415 Basin in the East Sea(Japan Sea) back-ark basin. Marine and Petroleum Geology, 18, 615- 416 634. 417 418 Lee, G.H., Yoon, Y., Nam, B.H., Lim, H., Kim, Y., Kim, H.J., and Lee, K., 2011, Structural 419 evolution of the southwestern margin of the Ulleung Basin, East Sea(Japan Sea) and 420 tectonic implications. Tectonophysics, 502, 293–307. 421 422 423 Lin, F., Ritzwoller, M.H., and Shapiro, N. M., 2006, Is ambient noise tomography across ocean basins possible?. Geophysical Research Letters, 33, L14304. 424 425 426 Liu X., Zhao D., and Li, S., 2013, Seismic imaging of the Southwest Japan arc from the Nankai trough to the Japan Sea. Physics of the Earth and Planetary Interiors, 216, 59–73. 427 19 428 Park G., Park J., Kwon B., and Park C., 2009, Tectonic structure modeling around the Ulleung 429 Basin and Dokdo using potential data. Journal of Korean Earth Science Society, 30, 165- 430 175. 431 432 Pawlak A., Eaton, D.W., Bastow, I.D., Kendall, J., Helffrich, G., Wookey, J., and Snyder, D., 433 2011, Crustal structure beneath Hudson Bay from ambient-noise tomography: implications 434 for basin formation. Geophysical Journal International. 184, 65–82. 435 436 437 Rawlinson, N., 2005, FMST: Fast Marching Surface Tomography package, Research School of Earth Sciences, Australian National University, Canberra, ACT 0200. 438 439 Rawlinson, N. and Sambridge M., 2005, The fast marching method: an effective tool for 440 tomographic imaging and tracking multiple phases in complex layered media. Exploration 441 Geophysics, 36, 341–350. 442 443 Sadeghi, H., Suzuki, S., and Takenaka, H., 2000, Tomographic low-velocity anomalies in the 444 uppermost mantle around the northeastern edge of Okinawa trough, the backarc of Kyushu. 445 Geophysical Research Letters, 27, 277–280. 446 447 Sato, T., Sato, T., Shinohara, M., Hino, R., Nishino, M., and Kanazawa, T., 2006, P-wave 448 velocity structure of the margin of the southeastern Tsushima Basin in the Japan Sea using 449 ocean bottom seismometers and airguns. Tectonophysics, 412, 159–171 450 451 Saygin, E. and Kennett, B. L. N., 2010, Ambient seismic noise tomography of Australian 20 452 continent. Tectonophysics, 481, 116–125 453 454 Schellart, W. and Lister, G., 2005, The role of the East Asian active margin in widespread 455 extensional and strike-slip deformation in East Asia. Journal of the Geological Society, 162, 456 959–972. 457 458 459 Seno, T., 1999, Syntheses of the regional stress fields of the Japanese islands. Island Arc, 8, 66–79. 460 461 462 Sethian, J., and Popovici, A., 1999, 3-D traveltime computation using the fast marching method. GEOPHYSICS, 64, 516–523. 463 464 465 Shapiro, N.M., and Campillo, M., 2004, Emergence of broadband Rayleigh waves from correlations of the ambient seismic noise. Geophysical Research Letters, 31, L07614, 466 467 468 Shapiro, N.M., Campillo, M., Stehly, L., and Ritzwoller, M.H., 2005, High-Resolustion Surface-Wave Tomography from Ambient Seismic Noise. Science, 307, 1615-1618 469 470 Son, M., Song, C.W., Kim, M., Cheon, Y., Jung, S., Cho, H., Kim, H., Kim, J.S., and Sohn, 471 Y.K., 2013, Miocene Crustal Deformation, Basin Development, and Tectonic Implication 472 in the Southeastern Korean Peninsula. Journal of the Geological Society of Korea, 49, 93- 473 118 474 475 Storchak, D.A., Di Giacomo, D., Bondár, I., Engdahl, E.R., Harris, J., Lee, W.H.K., Villaseñor, 21 476 A., and Bormann, P., 2013, Public Release of the ISC-GEM Global Instrumental 477 Earthquake Catalogue (1900-2009). Seismological Research Letter, 84, 810-815. 478 479 Tamaki, K., Suychiro, K., Allan, J., Ingle Jr., J.C., and Pisciotto, K.A., 1992, Tectonic synthesis 480 and implications of Japan Sea ODP Drilling. Proceedings of the Ocean Drilling Program, 481 Scientific Results, 127, 1333–1348. 482 483 484 van der Hilst R., Engdahl, R., Spakman, W., and Nolet, G., 1991, Tomographic imaging of subducted lithosphere below northwest Pacific island arcs. Nature, 353, 37-43. 485 486 Wei, W., Xu, J., Zhao, D., and Shi, Y., 2012, East Asia mantle tomography: New insight into 487 plate subduction and intraplate volcanism, Journal of Asian Earth Sciences, 60, 88–103. 488 489 Yoon, S.H., Sohn, Y.K., and Chough, S.K., 2014, Tectonic, sedimentary, and volcanic 490 evolution of a back-arc basin in the East Sea (Sea of Japan). Marine Geology, 352, 70–88. 491 492 Yoshizawa, K., Miyake, K., and Yomogida, K., 2010, 3D upper mantle structure beneath Japan 493 and its surrounding region from inter-station dispersion measurements of surface waves. 494 Physics of the Earth and Planetary Interiors, 183, 4–19. 495 496 Zhao, D., Maruyama S., and Omori, S., 2007, Mantle dynamics of western Pacific to East Asia: 497 New insight from seismic tomography and mineral physics. Gondwana Research, 11, 120- 498 131. 499 22 500 Zheng Y., Shen, W., Zhou, L., Yang, Y., Xie, Z., and Ritzwoller, M.H., 2011, Crust and 501 uppermost mantle beneath the North China Craton, northeastern China, and the Sea of Japan 502 from ambient noise tomography. Journal of Geophysical Research, 116, B12312. 503 504 505 23 506 Figures captions. 507 Fig 1. (a) Topographic map of the East Sea, South Sea and surrounding regions. Some intra- 508 volcanoes (yellow triangles) and many subduction zone volcanoes (red triangles) exist in this 509 area. The yellow contour indicates subduction depth of the Pacific plate and Philippine Sea 510 plate. Features: KP, Korean Peninsula; JI, Japan Island; ES, East Sea (Sea of Japan), KS, Korea 511 Strait; YS, Yellow Sea; NT, Nankai Trough; JT, Japan Trench; UB, Ulleung Basin; JB, Japan 512 Basin; YB, Yamato Basin; TI, Tsushima Island. (b) Location of KMA, KIGAM, F-net and GSN 513 stations (green triangles) and all possible ray paths (gray lines) used in this study. 514 515 Fig 2. Average group velocity for each period (white dot) is obtained from the average 516 Frequency Time Analysis (FTAN) diagram using the FTAN method. Amplitude of this figure 517 represents average amplitude of each FTAN diagram, normalized by the maximum amplitude. 518 The group velocity dispersion curve of the fundamental mode Rayleigh wave obtained from 519 the FTAN diagram is used as reference velocity for the 2-D group velocity map and to calculate 520 the sensitivity kernel for depth. 521 522 Fig 3. Group velocity maps (a–f) represent velocity perturbations from the average group 523 velocity for each period at 5, 10, 20, 25, 30, and 36 s. Significant high velocity anomalies are 524 clearly seen beneath the UB for periods between 20 and 30 s (Anomaly 1). Low velocity 525 anomalies surrounding the southern part of the basin are observed at 5 and 10 s (Anomaly 2). 526 Localized low velocity anomalies between TI and the UB appear at all periods (Anomaly 3). 527 Additionally, 1-D depth inversion is conducted at the white dots (1–4) that represent each 528 anomaly and a reference average. 529 24 530 Fig 4. Checkerboard test results with three different checker sizes (0.6°x0.6°, 1°x1° and 531 1.3°x1.3°). Results show that checkers with 0.6°x0.6° are well recovered in the KP and JI. 532 Recovered checker images of all checker sizes seem to be elongated in NW-SE direction 533 beneath the southern ES and KS. 534 535 Fig 5. Sensitivity kernel at 5, 10, 20, 25, 30 and 36s and 1-D structure inverted from the average 536 FTAN diagram. Rayleigh wave group velocities for periods of 5 and 36 s are sensitive to shear 537 wave velocities at depths of 5 and 40 km, respectively. 538 539 Fig 6. 1-D depth inversion results for points (1–4) that represent characteristic velocity 540 anomalies and the reference average. The black line represents the initial model of each point 541 and the inverted model is shown as a red line. The green line represents the inversion result of 542 model 4. The blue line indicates the inverted model with a 2-km thick water layer. 543 544 Fig 7. Group velocity map for the 5 s period and earthquake (2 ≤ M ≤ 6) epicenter locations 545 from 1994 to 2013 (ISC-GEM Catalog; Storchak et al., Seismological Research Letters, 2013). 546 Clearly, seismicity inside the UB is remarkably lower than surrounding regions. 547 25