3D Parallel Forward Modeling By The Fourier Transform Method

The project of seismic data forward modeling and poststack migration are studied in this dissertation. Seismic data migration, which is the key part of seismic data processing, is closely connected to the final purpose of seismic data processing and influence the final result of seismic exploration directly. Seismic data forward modeling is used through each step of acquisition, processing and interpretation. It is the foundation of seismic data migration and also the effective method for recognizing and researching the subsurface geologic structure.Phase-shift method is used for poststack migrating and forward modeling in K_x-wdomain. We can obtain migration result and forward modeling result by upgoing wave equation and exploding reflection model. We test migration followed by forward modeling and forward modeling followed by migration. The results prove that the migration and forward modeling are the reverse processes.The main difficulty in calculating the 3-D modeling is the requirement of the extremely large volume of computer memory and the high computational cost. The program is highly optimized in both computing efficiency and memory requirement by applying the vectorization and the relationship between the Fourier transform of complex sequence and that of real sequence to the 2-D and 3-D fast Fourier transform, which is the main computational part of the proposed algorithm. Theoretical analysis and results of model tests show that the frequency dispersion caused by approximating to the differential with the difference is reduced and better modeling result by adjusting the spatial and temporal sampling interval, the main frequency and length of the wavelet can be obtained. The Message Passing Interface (MPI) and PC-cluster are applied. It also helps to improve computing efficiency and save computing time. Combing the wave field snapshot with the corresponding velocity section proves to be important in the observing of wave propagation. The high quality synthetic seismic records obtained by the model test show that the algorithm is practical and efficient.In addition, the stacking velocity analysis based on the velocity spectra is described in the dissertation. The method is tested on the model datasets. And we get a horizontal layered velocity model that is analogous to the real velocity model.