Construction of a 3D Velocity Model of the Southern Korean Peninsula Using Ray-traveltime Forward Modeling

Accurate determination of earthquake source parameters requires that an assumed velocity model describe the medium as close to reality as possible. Therefore, a seismic survey with a dense linear receiver array using an active source such as an explosion is desirable to determine the accurate velocity structure. Observed traveltimes can be used in trial-and-error forward modeling based on the 2D ray-tracing approach to explore the velocity structure of the medium. The recent improvement of the Korean National Seismic Network (KNSN) in coverage and density enables permanent seismic profiling in any direction, similar to seismic exploration whenever natural or artificial earthquakes occur. We determined a 3D velocity model of the southern Korean Peninsula using the ray-traveltime forward modeling. We used seismic waveforms from earthquakes recorded on the KNSN between 2019 and 2020 for the forward modeling. Centering around the pre-determined hypocenter for each earthquake, a dozen of linear profiles consisting of the permanent seismic stations of the KNSN are omnidirectionally constructed with an equal spacing covering the whole surface. Then each profile of stations resembles the linear receiver array in the seismic exploration, which implies a layout of the 2D forward modeling. The initial model of the forward modeling consists of four layers representing the surface, upper and lower crust, and upper mantle, respectively, and each layer is discretized with small trapezoidal blocks. Ray tracing through the initial model is started, and then the calculated traveltimes of the P and S waves are obtained. The observed traveltimes are compared with the calculated ones. Then the model was repeatedly modified to minimize traveltime residuals which are differences between observed and computed traveltimes. In the final 3D model combining all the profiles, the depth of Moho discontinuity was the shallowest in the lower part of the Gyeonggi Massif and the deepest in the lower part of the Yeongnam Massif and the Gyeongsang Basin. The model shows that the depth to the upper/lower crust boundary is lower in the Gyeonggi Massif than in other regions. In some areas, a phase reflected from the intra-crustal boundary presumed to be the Conrad discontinuity is observed, and the depth to this boundary ranges between 14 and 20 km.