Cube2sph-GPU: A GPU accelerated toolkit enabling flexible continental-scale regional and teleseismic full waveform inversion

NANQIAO DU, Tianshi Liu, Bin He, Mijian Xu, Ting Lei, Qinya Liu

We present Cube2sph-GPU, a GPU-accelerated framework for continental-scale regional and teleseismic full-waveform inversion (FWI). Building upon the capabilities of SPECFEM3D_Cartesian, the toolkit introduces: (1) a flexible hybrid simulation scheme for tele-seismic simulations; (2) curvilinear C-PML; (3) spherical PDE-based kernel smoothing; and (4) highly optimized GPU kernels and I/O processing. The framework also includes a modular inversion engine supporting four distinct dataset types and multiple misfit functions. We validate the toolkit through two numerical experiments: an analysis of SKS wave sensitivity kernels and a multi-component ambient noise FWI for a Tilted Transverse Isotropy (TTI) model. These tests demonstrate the framework's ability to resolve complex anisotropic structures with high efficiency. Overall, Cube2sph-GPU provides a scalable and robust solution for high-resolution imaging of the Earth's interior.