Charles Carroll Burch

Prestack Depth Migration For Complex 2D Structure Using Phase-screen Propagators

We present results for the phase-screen propagator method applied to prestack depth migration of the Marmousi synthetic data set. The data were migrated as individual common-shot records and the resulting partial images were superposed to obtain the final complete Image. Tests were performed to determine the minimum number of frequency components required to achieve the best quality image and this in turn provided estimates of the minimum computing time. Running on a single processor SUN SPARC Ultra I, high quality images were obtained in as little as 8.7 CPU hours and adequate images were obtained in as little as 4.4 CPU hours. Different methods were tested for choosing the reference velocity used for the background phase-shift operation and for defining the slowness perturbation screens. Although the depths of some of the steeply dipping, high-contrast features were shifted slightly the overall image quality was fairly insensitive to the choice of the reference velocity. Our jests show the phase-screen method to be a reliable and fast algorithm for imaging complex geologic structures, at least for complex 2D synthetic data where the velocity model is known.

Hybrid local Born/Rytov Fourier migration method

Two efficient Fourier migration methods termed the extended local Born Fourier (ELBF) method and the extended Rytov Fourier (ELRF) method have been developed recently for imaging complex 3D structures. They are recursive methods based on local applications of Born and Rytov approximations within each extrapolation interval. The ELBF method becomes unreliable when the lateral slowness variations are large and/or the frequency is high, while the ELRF method is reliable for such cases. However, the ELRF method is approximately 30-40% slower than the ELBF method because the ELRF method requires one more computational step where exponentials of complex numbers are calculated than the ELBF method and propose an implementation scheme using variable extrapolation intervals to make the ELBF method reliable for all lateral slowness variations and frequencies. The size of the extrapolation interval depends on the lateral slowness variations within a given extrapolation region and the frequency, and consequently, the computational time of the ELBF method with variable extrapolation intervals increases with the lateral slowness variation and frequency. To take advantage of the faster computational speed of the ELBF method compared to the ELRF method and the better stability of the ELRF method compared to the ELBF method, we propose a hybrid local Born/Rytov Fourier migration method. In the hybrid method, the ELBF method is used for regions with small lateral slowness variations and/or low frequencies, otherwise, the ELRF method is used. Migrations of two synthetic datasets for complex structures using the ELBF method with variable extrapolation intervals and the hybrid method demonstrate that the quality of images obtained using these two methods is comparable to that of images obtained using the ELRF method. Comparison of computational times for migrations using different methods shows that the ELBF method with variable extrapolation intervals takes much more computational time than the ELRF method but the hybrid method saves more than 10% of the computational time required by the ELRF method.

Practical aspects of prestack depth migration with finite differences

Finite-difference, prestack, depth migrations offers significant improvements over Kirchhoff methods in imaging near or under salt structures. The authors have implemented a finite-difference prestack depth migration algorithm for use on massively parallel computers which is discussed. The image quality of the finite-difference scheme has been investigated and suggested improvements are discussed. In this presentation, the authors discuss an implicit finite difference migration code, called Salvo, that has been developed through an ACTI (Advanced Computational Technology Initiative) joint project. This code is designed to be efficient on a variety of massively parallel computers. It takes advantage of both frequency and spatial parallelism as well as the use of nodes dedicated to data input/output (I/O). Besides giving an overview of the finite-difference algorithm and some of the parallelism techniques used, migration results using both Kirchhoff and finite-difference migration will be presented and compared. The authors start out with a very simple Cartoon model where one can intuitively see the multiple travel paths and some of the potential problems that will be encountered with Kirchhoff migration. More complex synthetic models as well as results from actual seismic data from the Gulf of Mexico will be shown.