1-D velocity model for the North Korean Peninsula from Rayleigh wave dispersion of ambient noise cross-correlations
- 134 Downloads
Monitoring seismic activity in the north Korean Peninsula (NKP) is important not only for understanding the characteristics of tectonic earthquakes but also for monitoring anthropogenic seismic events. To more effectively investigate seismic properties, reliable seismic velocity models are essential. However, the seismic velocity structures of the region have not been well constrained due to a lack of available seismic data. This study presents 1-D velocity models for both the inland and offshore (western East Sea) of the NKP. We constrained the models based on the results of a Bayesian inversion process using Rayleigh wave dispersion data, which were measured from ambient noise cross-correlations between stations in the southern Korean Peninsula and northeast China. The proposed models were evaluated by performing full moment tensor inversion for the 2013 Democratic People’s Republic of Korea (DPRK) nuclear test. Using the composite model consisting of both inland and offshore models resulted in consistently higher goodness of fit to observed waveforms than previous models. This indicates that seismic monitoring can be improved by using the proposed models, which resolve propagation effects along different paths in the NKP region.
KeywordsSeismic velocity structure Bayesian inversion Rayleigh wave dispersion North Korea Peninsula Nuclear test
A reliable seismic velocity model is essential for monitoring and analyzing seismic activity and associate hazards. In addition, a well-determined seismic velocity model provides basic information for understanding the tectonic process of a region. However, seismic velocity structures in the northern Korean Peninsula (NKP) region in northeast Asia are not well studied due to a lack of available seismic networks. In contrast, dense seismic networks have been installed in the rest of northeast Asia, including northeast China, the southern Korean Peninsula (SKP), and the Japanese islands, and seismic velocity structures are therefore extensively mapped in these regions (e.g., Zheng et al. 2011; Wei et al. 2012; Witek et al. 2014; Shen et al. 2016; Kim et al. 2016b). In particular, one-dimensional (1-D) structures of the SKP have been estimated from studies using receiver functions and seismic waveform data (Chang and Baag 2005; Chang and Baag 2006; Lee and Baag 2008; Han et al. 2010; Kim et al. 2011; Kim et al. 2016a). Studies of the NKP region (e.g., Shin and Baag 2000; Ford et al. 2010) have relied on approximate velocity structures based on the SKP or global averages.
Despite relatively low seismicity, monitoring of seismic activity in the NKP region is particularly important. The largest instrumental earthquake (Mw 6.2) in the Korean Peninsula occurred near Pyongyang in North Korea, where more historical records of significant earthquakes are found (Kang and Jun 2011; Kyung et al. 2016). In addition, six nuclear explosion tests have been conducted since 2006 by the Democratic People’s Republic of Korea (DPRK) government, generating the need for effective monitoring and detection of future potential activity (Hong and Rhie 2009; Barth 2014; Cesca et al. 2017).
In regions with low seismicity or a lack of instrumentation, such as the NKP region, cross-correlation of seismic ambient noise can be utilized to obtain Green’s functions between station pairs (Shapiro et al. 2005; Bensen et al. 2007). The method is widely used for investigating velocity structures of the crust and upper mantle because it is effective for obtaining shorter period data of crustal depth than earthquake data (e.g., Bensen et al. 2007; Kang and Shin 2006; Choi et al. 2009; Pawlak et al. 2011; Shen et al. 2016; Kim et al. 2016b). In this study, we developed 1-D velocity models for the NKP region using Rayleigh wave dispersion data obtained from ambient noise cross-correlations between permanent network stations in SKP and a temporary network in northeast China (Northeast China Extended Seismic Array, NECESSArray). Two different models were presented to account for different propagation paths along the inland and offshore regions of the Korean Peninsula. The reliability of the estimated 1-D models was verified by calculating full model tensor solutions for the 2013 North Korea explosion nuclear test.
2 Data and methods
2.1 Data and processing
2.2 Bayesian inversion
The obtained surface wave dispersion data were inverted to estimate the 1-D shear wave velocity models for the corresponding path coverage regions. It is known that surface wave dispersion data have broader sensitivities to shear wave velocity structures than ballistic wave data such as body wave travel times (Ritzwoller et al. 2011; Obrebski et al. 2011). In this study, therefore, we utilized a Bayesian inversion method to account for potential non-uniqueness of solutions by estimating uncertainties, which were deduced from posterior probability distributions (PPD) of inversion parameters (Bodin et al. 2012; Kim et al. 2016a; Kim et al. 2016b). In particular, robust inversion uncertainties were calculated through a recently developed trans-dimensional and hierarchical technique (see Kim et al. 2016a and Kim et al. 2017 for further details). In this approach, probability distributions of the number of layers and the levels of data errors (variances) are sampled, through a Markov chain Monte Carlo method, together with those of other inversion parameters (e.g., layer thicknesses and velocities in each layer). Without applying regularization in the inversion process, this scheme systematically controls the balance between model complexity (i.e., the number of layers) and the degree of data fitting (i.e., the level of data errors). As a consequence, inversion results and corresponding uncertainties are less biased by arbitrary inversion settings. In addition, the approach is particularly useful in joint inversions as group and phase velocity dispersions are used together. The hyper-parameters for implementing the hierarchical scheme encompass the effect of different data sensitivities between different types (group and phase velocity) of data.
We used Rayleigh wave group and phase velocities from the average of the dispersion curves for each model (3~50 s-period for Mod_Land; 3~45 s-period for Mod_Sea). Bayesian inversions were performed for parameters consisting of Vs (2.0 ≤ Vs ≤ 5.5 km/s), Vp/Vs ratio (1.7 ≤ Vp/Vs ≤ 2.0), interface depth (up to 100 km), number of layers (between 2 and 30), and scaling factors for the level of data noise (between − 5.0 and 5.0). In each inversion, 2000 iterations were progressed and the previous 1000 samples were discarded to obtain converged posterior models. We simultaneously used 64 parallel chains to finally form a mean model from 1000 posterior sampled models. Figure 4 shows the prior ranges and estimated PPD for two different datasets. Based on the assumption that the PPD can be approximated as Gaussian shapes, we present inversion uncertainties with ± 2 standard deviations of the sampled models.
3 Results and discussion
Model parameters of Mod_Land and Mod_Sea. L1 and L2, bottom depth of 1st and 2nd layer; S1–S3, S wave velocity of each layer; and P1–P3, P wave velocity of each layer
6.5 ± 1.0 (±1σ)
30.0 ± 1.1
3.36 ± 0.03
3.57 ± 0.07
4.34 ± 0.08
5.82 ± 0.05
6.22 ± 0.1
7.64 ± 0.13
7.0 ± 1.0
25.0 ± 1.2
3.34 ± 0.02
3.51 ± 0.09
4.27 ± 0.1
5.78 ± 0.04
6.11 ± 0.13
7.52 ± 0.17
We qualitatively evaluate the obtained models by comparing with previous 1-D velocity models in the SKP. The first is a model determined through travel time data and waveform modeling of seismic phases of local earthquakes (Chang and Baag 2006; Mod_C&B). The second is a model that best fits the observation waveforms determined by a full-grid search (Kim et al. 2011; Mod_Kim). Mod_Land is generally similar to Mod_C&B and Mod_Kim, even taking into account the different number of layers (four) (Fig. 5). The Moho depth in Mod_Land developed using inland sampling data is similar (~ 30 km) to those in Mod_C&B and Mod_Kim and the velocity difference of each layer of Mod_C&B and Mod_Kim compared with Mod_Land is less than 4%, except for the lower crust part. It is known that the lower crust has been significantly modified by tectonic processes (e.g., Chough et al. 2000) in each region of the Korean Peninsula.
The influence of sampling paths through northeast China on Mod_Land is accounted for by the portion of inter-station paths that crosses the region. Considering previous studies on velocity structure in northeast Asia, the average crustal velocity in northeast China (~ 3.6 km/s) is not significantly different from that in the Korean peninsula (Zheng et al. 2011; Shen et al. 2016). In Songliao basin, the uppermost crustal velocity is slower than in the Korean peninsula by approximately 10% due to thick sediments. The Moho thickness is also relatively shallow; less than 30 km (Guo et al. 2015; Shen et al. 2016). However, the stations selected in this study are located at the boundary of the Songliao basin and the portion of paths through the basin does not exceed 20% of their total path-lengths. In addition, the standard deviation of the average dispersion curve for Mod_Land is small, even though it used paths across various regions (Fig. 4). Consequently, we conclude that paths are less influenced by the northeast China region and Mod_Land is sufficiently representative of NKP.
In Mod_Sea, the depth of the Moho is approximately 25 km, which is relatively shallow compared with Mod_Land and South models. Moreover, the velocity difference from Mod_Land is ~ 2% in each layer. This relatively small discrepancy indicates that, as previously suggested, the eastern margin of NKP is not completely oceanic crust but a rifted continental margin formed during expansion (Tamaki 1988; Yoon et al. 2014).
3.2 Model validation
The FMT inversion results show that the Mod_Land and Mod_Sea models can improve the seismic monitoring of DPRK nuclear tests. Furthermore, this work indicates that the effect of wave propagations should be considered; developing more refined path-specific 1-D models or higher-dimensional models may further improve the FMT inversion results. In the future, radially anisotropic models can also be developed and tested because this study only assumes isotropic structures based on Rayleigh wave dispersion data.
In this study, we develop 1-D velocity models representing the inland and eastern margin of the NKP through the Bayesian inversion method using dispersion curves from the ambient noise records of surrounding networks (Figs. 4 and 5, Table 1). Although the velocity difference is small between Mod_Land and Mod_Sea models, Moho depths are clearly identified at 30 and 25 km, respectively. Mod_Land is similar to previous 1-D models (Mod_C&B and Mod_Kim) in the SKP region, which indicates that continental velocity structures in the Korean Peninsula are generally similar. Model verification using the FMT inversion for the 2013 DPRK nuclear test shows that a composite set of models consisting of Mod_Land and Mod_Sea for paths along the inland and eastern margin of NKP, respectively, provide a better solution in terms of the goodness of fit (VR) to waveforms than cases using either one of the models or with any possible combination of AK135 and Mod_Kim. Additionally, we confirm the reliability of the composite model, which may lead to improved FMT estimation for seismic events in NKP.
This work was funded by the Korea Meteorological Administration under grant KMIPA2017-4010.
- Ford SR, Dreger DS, Walter WR (2009) Source analysis of the Memorial Day explosion, Kimchaek, North Korea. Geophysical Research Letters 36 (21)Google Scholar
- Herrmann RB, Ammon CJ (2004) Surface waves, receiver functions and crustal structure. Computer programs in seismology: Version 3.30. Saint Louis UniversityGoogle Scholar
- Hong TK, Baag CE, Choi H, Sheen DH (2008) Regional seismic observations of the 9 October 2006 underground nuclear explosion in North Korea and the influence of crustal structure on regional phases. J Geophys Res Solid Earth 113(B3)Google Scholar
- Kang TS, Jun MS (2011) Some studies on the 1952 earthquake near Pyeongyang, North Korea, The 1st Annual Meeting of the Project of Seismic Hazard Assessment for the Next Generation Map: Strategic Japanese-Chinese-Korean Cooperative Program: Seismic Hazard Assessment for the Next Generation Map, November 25-30, 2011, Harbin, China, http://www.j-shis.bosai.go.jp/intl/cjk/1st_annual_meeting.html (last accessed on June 8, 2017)
- Kang TS, Shin JS (2006) Surface-wave tomography from ambient seismic noise of accelerograph networks in southern Korea. Geophysical Research Letters 33 (17)Google Scholar
- Tamaki, K (1988) Geological structure of the Japan Sea and its tectonic implications. Bulletin of the Geological Survey of Japan, 39, 269–365Google Scholar
- Zheng Y, Shen W, Zhou L, Yang Y, Xie Z, Ritzwoller MH (2011) Crust and uppermost mantle beneath the North China Craton, northeastern China, and the Sea of Japan from ambient noise tomography. Journal of Geophysical Research 116 (B12)Google Scholar
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.