University of Texas, Austin
Seismic data are used to infer subsurface structures and reservoir properties. Current data acquisition techniques employ 12 or more receiving arrays of up to 6.0 km in length with sources of seismic energy activated every 50m or less. These 3D surveying methods challenge traditional seismic data processing and imaging methods in at least two ways.
First, the data volumes recorded require computationally efficient yet effective algorithms. Second, the excellent survey coverage and the desire for accurate subsurface images make wave equation based depth imaging not just possible but a requirement. To achieve this goal, the seismic propagation velocity must be derived from the recorded data as part of the imaging process. These challenges can be addressed by decomposing the recorded data into its plane wave constituents. This reduces the data volume since appropriate subsets of the recorded information can be used in the imaging and velocity analysis.
In addition, many triplications observed in conventional offset-time domain are unraveled in the plane wave data. Classical integral methods for seismic imaging, known as Kirchhoff migration, have plane wave analogs that are computationally efficient and accurate. Plane wave frequency-wavenumber perturbation methods can be formulated that have better resolving power when lateral velocity variations are small. Finite-difference approximations can also be employed. Velocity estimation is a non-linear optimization problem that is well suited for directed Monte Carlo methods such as very fast simulated annealing.
By evaluating the coherency of variable incidence angle image estimates for
each subsurface point, an objective function can be formed and used to update
the velocity model. Rapid imaging algorithms acting on selected subsets of plane
wave data are one approach to solving the velocity estimation problem, which
is a prerequisite for accurate 3D depth imaging.