Least‐squares reverse time migration with flux‐corrected transport technique

Abstract

AbstractDue to high computational costs and memory requirements, it is necessary to use coarse grids for least‐squares reverse time migration when processing large‐scale field data. According to the dispersion relation, strong numerical dispersion will occur in least‐squares reverse time migration images when the migration velocity of the shallow strata is low, reducing the resolution of the migration results. To overcome this shortcoming, we propose new least‐squares reverse time migration implementation method with a modified flux‐corrected transport technique. This algorithm involves two main steps: obtain an appropriate calibration parameter for flux‐corrected transport through forward modelling and perform least‐squares reverse time migration using flux‐corrected transport. We remove the non‐linear processing from conventional flux‐corrected transport algorithm often used in computational fluid dynamics based on the assumption that the seismic wavefield is relatively continuous and smooth without strong ripples, which makes it possible to apply flux‐corrected transport to least‐squares reverse time migration without increasing too much computation complexity. To suppress the numerical dispersion more effectively, the diffusion fluxes along the diagonal direction are added to flux‐corrected transport. Numerical tests on synthetic data and field data illustrate that compared to the conventional least‐squares reverse time migration method, the proposed method produces images with fewer migration artefacts, more accurate amplitudes and faster convergence speed on the coarse grids.