Reverse Time Migration And Constructing Angle Gathers

Reverse time migration (RTM) is a migration method with high imaging accuracyand no dip limitaion. It can be used when the subsurface structuresare complex, and isable to image return waves, multiples, prism waves. However, it also has its owndisadvantages: huge demand of storage space, large amount of computation, and seriouslow-frequency artifacts. An accurate velocity model is a prerequisite to obtain accurateimages, and angle domain common image gathers (ADCIGs) provides data for velocityanalysis. This paper performs research on RTM mainly from the following aspects are:(1) The effects of finite difference orders, grid spacing, the main frequency of asource wavelet, model velocities on dispersion. Tests show that dispersion phenomenonis weakened with increasing of the finite difference orders, decreasing of grid spacingand the main frequency of the source wavelet. A high velocity can also suppressdispersion. However, the main frequency and velocities should be determined accordingto true situation in the field.(2) Boundary conditions in wavefields propagation. Different absorbing boundaryconditions are compared and the Perfectly Matched Layer absorbing boundary conditionhas a better effect of absorbing bound reflections. In the paper, to reduce the demand ofstorage space, a stochastic boundary condition is used to ensure that that sourcewavefields can be reversed backward from the maximum time, as well as reduce thecorrelation among the boundary reflections in both source wavefields and receiverwavefileds.(3) GPU/CPU heterogeneous parallel computation. Multi-threads parallelcomputation and GPU storage optimization are mainly used to accelerate RTM andADCIGs computation.(4) Migration depths of reflectors in RTM. Operation of RTM causes a change of45degree in the phases of wavelets in seismic data, which affects the depths of eventsin migration images. Combined with phase correction, the forward propagating time ofsource wavefields is lengthened by half of the length of the source wavelet to gainedcorrect migration depths. (5) Constructing ADCIGs using Poynting Vectors. The concept of PoyntingVector was first used in electromagnetism, representing the direction of main energy ofthe wavefileds. In RTM, one order directives to space and to time of the wavefields arecalculated first, then Poynting Vectors are. Reflection angles between source wavefiledsand receiver fields are computed after knowing the Poynting Vectors. To ensure thestability and a high signal to noise ratio in ADCIGs, smoothing, multi-time stepsstacking of Poynting Vectors and Gauss sample function are performed in theprocedure.(6) Imaging condtions and artifacts suppressing. In RTM, low-frequency artifactsare mainly produced when the source wavefields and receiver wavefields propagate inopposite directions, that is, artifacts mainly focus on large angles in ADCIGs. Thensuppressing images in large angles can achieve the effect of suppressing low-frequencyartifacts. The methods of Laplace domain filtering and cross-correlation imagingcondition with Poynting Vectors are based on the theory mentioned above.