A new generation of depth imaging algorithms, two orders of magnitude faster than the standard Kirchhoff depth migration, has arrived. These clever algorithms, combined with very fast migration velocity analysis tools, can reduce the turnaround time for large 3-D seismic projects from six to eight months down to one to two months.

These "Smart Migrations" algorithms use the information inside the prestack data for migration as opposed to standard depth migration algorithms like Kirchhoff migration or reverse time migration (RTM), which assume every image point in the subsurface is a diffractor and make no a priori assumptions about the migrated image structure. Instead, Smart Migrations analyze the input data, build a database of locally coherent events, and use these events to guide the direction of the migration operator. This knowledge of the input data can speed up the processing time by 100 to 500 times.

Processing cycles for high-resolution subsurface imaging can take six to eight months for medium to large depth imaging projects in the 3-D seismic services industry and two to three years in the internal processing centers of some of the major oil companies. While the computer industry is increasing the computer speed and efficiency according to Moore's law, doubling CPU speed every 18 months, the seismic industry is matching and exceeding the computational capabilities of the processing centers by acquiring higher resolution, higher density wide-azimuth datasets and using more complex and more compute-intensive processing algorithms like RTM.

According to Larry Lunardi, vice president of geophysics at Chesapeake Energy Corp., "With today's high-density, wide-azimuth seismic recording capabilities, even a relatively small-scale exploration program can generate enormous volumes of raw 3-D data that require massive amounts of compute power to acquire, store, manage, process, visualize, and interpret."

This image shows the ratio between the extent of an impulse response in Kirchhoff migration and the much reduced extent of a beam. (Images courtesy of Z-Terra Inc.)

Large-input 3-D datasets are typically tens to hundreds of terabytes in size. During processing the size can expand to thousands or tens of thousands of terabytes through phase-space unrolling or by generating extended image gathers. Processing such large datasets requires better algorithms, adequate visualization, and quality control.

Obtaining an image of the subsurface requires solving an inverse problem for the wave equation in 3-D. Efficiency requires that these processes be monitored in real time and that intermediate output be visualized and analyzed throughout the process.

Ten years ago the seismic processing industry had data density of thousands of traces per square mile; now some of the surveys can have millions of traces per square mile. Ten years ago the industry had thousands of channels; now companies announce records of 45,000 shot points per day and 100,000 channels. Counting the bytes, 45,000 shot points multiplied by 100,000 channels multiplied by 10 seconds per trace multiplied by 250 samples per second at 4 bytes per sample give densities of 45 terabytes of data per day. These are record numbers and give a preview of what will be common in five to 10 years. Also working against Moore's Law is the ever-increasing algorithm complexity, requiring more and more computer resources. This drives an ever-increasing need for computation power. On this background of increasing data density and algorithm complexity, Smart Migrations are a great tool to have in the arsenal of the imaging groups.

Smart Migrations

The Smart Migrations class of algorithms is a perfect response to the industry need to handle large amounts of data in a short time. One early and successful commercial implementation of a Smart Migration class algorithm is the Fast Beam Migration (FBM) developed by John Sherwood at Applied Geophysical Services. The speed of FBM is achieved in two steps: 1. A factor of 10 to 100 in speedup is achieved using beam forming, or beam decomposition of the input data, where the number of input data is reduced by a factor of 10 to 100. 2. A factor of 10 to 100 in speedup is obtained by spreading each input trace or beam over a beam instead of a full aperture-volume. The Z-Terra Fast Beam Migration (FBM) implementation follows a similar architecture: 1. Beam forming is performed by using automatic plane-wave destruction, plane-wave construction, and shaping. 2. Beam extrapolation and imaging is performed derived from constructions employed previously in wavepath and parsimonious migration and oriented imaging. The first step of the FBM decomposes the seismic data in wavelets, 100 to 200 ms in time duration. The beam formation is done in D(t,x,y,h) coordinates for common azimuth data and D(t,x,y,h,a) for wide-azimuth data, or directly in D(t,Sx,Sy,Gx,Gy) coordinates, where x ,y, h, and a are common data point inline and crossline coordinates, offset and azimuth, and Sx, Sy, Gx, and Gy are shot and receiver X,Y coordinates. The beam formation reduces the input data volume by a factor of 10 to 100 and is an important factor in the speedup achieved with FBM. For larger reduction factors, data are not invertible; the original data cannot be fully recovered from the data beam. But at a larger data compression rate, the beams increase the signal-to-noise ratio and retain the coherent energy in the data. The migrated image appears almost free of artifacts due to this signal-to-noise enhancement and makes the image easier to interpret in complex areas.

The image on the left shows the extent of Kirchhoff impulse response. On the right is the reduced extent of a beam.

The faster imaging step allows for more iterations of velocity model building (30 to 40 iterations, instead of the current seven to 10), which enables the processing team to enhance the seismic resolution and imaging of complex geologic structures and allows for deeper data, steeper dip, and subsalt structure imaging. The velocity model is critical to the dataset's ability to accurately image the complex allochthonous salt canopies floating above the deeper Lower Tertiary structures, the overthrust sheets in tectonically deformed and folded areas, or fault shadow areas where the velocity model varies laterally due to complex systems of faults. In the presence of these complex geometries, the seismic energy propagating down and reflected back from the underlying sediments experiences very high signal distortion and dispersion. Combining accurate and fast velocity model-building with advanced imaging methodology improves the success rate and cost effectiveness for new deep-field discoveries, reduces the turnaround time for large surveys, and also has applications in increasing recovery efficiency for the development of existing fields. This technology does not exist widely in the industry yet, but it is a fundamental advance and necessary building block in any seismic processing system that uses depth processing methods for imaging deep and complex geological structures, the focus of modern oil and gas exploration.

References available.