Finite-difference modeling with a minimal computation cost

Abstract

Summary I develop a time-space finite-difference (FD) modeling scheme with a minimal computation cost. For the same velocity model, temporal length and modeling accuracy, when variation of computation cost with grid size is obtained, one may find a minimal computation cost, and thus the corresponding modeling parameters such as grid size, time step and spatial FD operator length may be determined. The key is to automatically determine maximum time step and minimum spatial FD operator length. Recently developed time-space domain dispersionrelation-based finite-difference method and variable spatial FD operator lengths method make it possible and feasible. Examples for 2D acoustic FD modeling show that a minimal computation cost can be found and CPU time of FD modeling demonstrates the feasibility and correctness of the proposed scheme.