Ambient noise tomography, when applied to a dense linear seismic array, has the capability to provide detailed insights into the fine velocity structures across diverse tectonic settings. The linear station arrangement naturally generates parallel and concentrated ray paths along the array trend. This unique geometry requires specific optimization of the inversion methodology and model parameterization. The Bayesian-based transdimensional inversion method, characterized by its fully non-linear nature and high degree of freedom in parameter settings, offers a powerful tool for ambient noise inversion. To effectively adapt this method to a linear array layout, we propose a modification to the Voronoi cell tessellation built in the transdimensional method. By introducing spatial priority to the Voronoi kernels, we strategically increased the density of Voronoi cells along the direction of the array. We then applied the modified approach to a linear seismic array in the North China Craton and validated its robustness through phase velocity images and resolution tests. Our improved non-uniform sampling technique in the 2-D model space accelerates convergence while simultaneously enhancing model accuracy. Compared with the conventional damped least-squares method, the proposed algorithm revealed a shear-wave velocity map with notable low-velocity anomalies situated in the middle and lower crust beneath the borders of the Ordos block and its surrounding orogenic belt. Aligned with the crustal structures revealed by receiver function and electrical imaging, our findings indicated that the western and eastern margins of the Ordos block had experienced intensive crustal wedge deformation and re-melting, respectively.
Read full abstract