Abstract

The Spectral-Element Method (SEM) offers a high-accuracy alternative to conventional FiniteDifference (FD) approaches for seismic modelling and inversion. FD methods are computationally efficient on quasi-regular Cartesian grids, but offer diminished accuracy for complex geological interfaces (e.g. the surface topography) and when dealing with coupleddomain fluid/solid boundary conditions. In contrast, SEM employs unstructured meshes and high-order polynomial approximations, enabling it to accurately model waveforms in the presence of complex topography, rugose bathymetry and heterogeneous subsurface structures. With the integration of advanced numerical techniques and GPU acceleration, SEM delivers precise full-waveform modelling and inversion capabilities suited for challenging geophysical scenarios. Field data from offshore Western Australia highlights SEM’s ability to accurately discretize intricate seabed geometries and delivers improved results for both waveform modelling and inversion when compared to FD methods.