Abstract:The paper presented split-step Fourier prestack depth migration method using velocity structure to partition the velocity field.The theoretical basis of Split-Step Fourier (SSF) method is disturbance item,when the velocity contract is large,the errors of disturbance item become larger in one step length of wavefield continuation.Therefore,when lateral velocity has rapid variation,the velocity partition is implemented according to velocity structure and using different reference velocities in each velocity zone,which can greatly reduce the contracts between the reference velocities and real velocities and then increase the imaging precision of split-step method;meanwhile,it can greatly reduce the numbers of used reference velocities and reduce the computational efforts to minimum.The implemented steps are follows:①the max and min velocities of a particular layer are computed within each continuation step first,then an adequate boundary velocity is defined between both velocities;②taken the boundary velocity as boundary, the velocity zone is divided into several high-velocity sub-zonesand low-velocity sub-zones and the velocity continuity between different sub-zones is unnecessary;③using different reference velocities in different velocity sub-zones to implement wavefield continuation by SSF method,i.e.phase-shift is implemented first in the frequency-wavenumber domain,then split-step method is realized point by point back to the frequency-space domain;④merging the different wavefields resulted from partitioned continuation according to relevant grid points and carrying out the adequate smooth processing when different wavefields are merged;⑤repeat abovementioned steps until finishing continuation and mergence of all velocity partitions.Theoretical analysis,poststack depth migration of simple models and prestack depth migration of Marmousi model showed that the method is feasibility.