A Second-order Parallel Fast Sweeping Method for Traveltime Calculation

Summary Traveltime computation is fundamental for seismic data analysis, impacting the quality of migration images and near-surface velocity models. As exploration targets become more complex and seismic datasets expand, the demand for precise and efficient traveltime calculations has increased. This paper introduces a second-order parallel fast sweeping method (FSM) that addresses these needs by solving Eikonal equations on GPUs with enhanced accuracy and efficiency. The proposed FSM utilizes the Cuthill-McKee algorithm alongside alternating direction sweepings to achieve this. It employs a two-point stencil for second-order accuracy when two upwind points are available; otherwise, it defaults to first-order. The second-order scheme offers superior solutions over first-order methods, achieving the same level of accuracy with fewer iterations and less memory usage. Importantly, the method’s parallelization on GPU significantly accelerates computations for large 3D cases. Unlike the first-order approach, reducing grid size does not sufficiently mitigate errors, which can impact migration velocity analysis. Thus, the second-order FSM provides a robust solution for high-resolution seismic imaging and velocity model building.