Adjoint Full-Waveform Tomography in Taiwan

To inverstigate the subsurface structures beneath Taiwan, numerous tomographic models have been developed over the last few decades. However, all of them are ray-based tomography and rely on simplified assumptions. In this study, we aim to improve the velocity model across the whole Taiwan region by leveraging full-waveform information, which provides stronger constraints than traditional travel-time data.  So far, we have enhanced the velocity model within the 2-30 second period band. 

Studying the Causes of Double P Phases observed around Northeastern Taiwan

Previous studies observed dual P phases from the signals of several deep events beneath Northeastern Taiwan. They identified that the anomalous observations are caused by the scattering from magma diapirs on the basis of a 1D layered model and ray-tracing method. However, waveform modeling based on the model with heterogeneities is necessary for explaining this complicated path effect. Here, we construct the velocity models on the basis of previous tomographic studies. Next, we simulate the waveforms using these self-created models, in order to find the model with the smallest misfit. According to our result, the high-velocity zone below Northeastern Taiwan and the interface between the mantle wedge and the high-velocity Eurasian lithosphere might be the main causes for producing the double P phases. (Yen et al. (2025), PEPI )

Validating the Joint inverted tomographic model using numerical waveform modeling

In our group, we have developed a high-resolution seismic imaging around Yilan, Taiwan using the data of body wave, surface wave, and ambient noise. Here, we utilize the numerical waveform modeling to assess the reliability of this velocity model. By placing a single-force source slightly beneath a specific station, we can extract the Green's function between every station pair. Therefore, the similarity between observational and synthetic Green's function demonstrates the robustness of incorporating ambient noise data into the joint inversion.