[S-SS05] Advancements in Regional Seismic Networks: Operations, Applications, and Development

[SSS05-P04] Multi-Mode Ambient Noise Phase Velocity Tomography in Southern Taiwan Using Frequency-Bessel Transformation and Voronoi Partitioning

*Wen-Hao Wu¹, Shu-Huei Hung¹, Jun-Fu Liao², Yin-Nien Chen², SALUTE Team -^1,2,3,4 (1.Department of Geosciences, National Taiwan University, 2.Department of Earth and Environmental Sciences, National Chung Cheng University, 3.Department of Earth Sciences, National Taiwan Normal University, 4.Institute of Earth Sciences, Academia Sinica)

Keywords:Ambient Noise Cross Correlation, Frequency-Bessel Transformation, Voronoi Partitioning, Multi-mode Surface Wave Velocity, Machine Learning

Phase velocity dispersion derived from the cross-correlation functions (CCFs) of ambient seismic noise between station pairs has been proven effective for tomographic imaging of subsurface structures. Traditionally, most studies rely on the fundamental mode of surface waves, which suffer from poor depth resolution and nonuniqueness in shear wave velocity inversion due to its smooth and depth-decaying sensitivity kernel. To address these limitations, this study employs a multi-mode approach to surface wave phase velocity tomography in southern Taiwan, using data from the dense broadband SALUTE array deployed across southern Taiwan and eastern offshore region since October 2021. This region represents a key transition from continental subduction to arc collision, making high-resolution tomographic imaging essential for understanding its tectonic evolution. We implement a Voronoi partitioning approach to randomly and repeatedly divide the study area into multiple subregions. Within each subregion, the Frequency-Bessel (F-J) transformation is applied to ambient noise CCFs to generate the F-J spectrogram for multimode phase velocity dispersion measurement. A machine-learning clustering method is then used to automatically extract both fundamental and higher-mode dispersion curves. By generating 200 realizations of the Voronoi diagram, we construct frequency-dependent phase velocity maps and quantify their uncertainty at each target location by averaging all the determined phase velocities sampled across partitioned Voronoi cells. This multi-mode Voronoi-based tomography significantly improves depth sensitivity and enhances structural resolution, potentially resolving more detailed shear wave velocity structures in the crust and lithospheric mantle beneath southern Taiwan.