Accelerating anisotropic elastic subsurface imaging with dual-pair finite differences

Anisotropic elastic full waveform inversion based on finite difference modeling on fully staggered grids like the Lebedev grid comes with a high computational cost. We examine using a non-staggered finite difference method instead, as a possibility to accelerate elastic subsurface imaging based on fully staggered grids. Most existing studies on finite difference elastic modeling on non-staggered grids are limited to modeling waves sampled above 4 points per P wavelength. This makes them relatively inefficient for subsurface imaging using P waves, compared to the Lebedev grid which can accurately model waves sampled below 3 points per P wavelength for a suitable stencil width. We show that accurate finite difference wave modeling on non-staggered grids below 3 points per P wavelength can be achieved by a combination of dual pairing,selective filtering, and some adjustments to source injection. Preliminary tests on field data show that the proposed methodology can lead to roughly 3 times faster subsurface elastic imaging using P waves, compared to using the Lebedev grid for the same sampling rate, even below 3 points per P wavelengths.