Multiple reflections as an additive noise limitation in seismic reflection work

A principal problem in seismic exploration is the heterogeneous partially reflecting medium that makes up the transmission zone to the target reflectors of interest. Multiple reflection between the free surface and the shallower reflectors produces additive noise that interferes with our view of the bedding plane geometry and rock properties in the target zone. Residual multiple energy on the interpreter's record section display can easily be mistaken for signal. Because the magnitude of multiples is dependent on reflectivity products, and because subsurface reflectivities are normally small, the importance of multiples is highly variable. The complementary combination of predictive deconvolution and common depth point stack is routinely used for the reduction of multiple reflections. Predictive deconvolution achieves a temporal prediction and subtraction of that class of short-period multiples that is most predictable, and common depth point stack reduces that class of multiples distinguishable from primaries on the basis of local horizontal phase velocity in the x, t observational space. Each of these methods can be justified in terms of a simplified one-dimensional model of the earth. However, a major reason for their success is that they are employed with an empirical, statistical approach. As a result, these two processes are particularly robust and forgiving and respond imperfectly, but often constructively, to real-world deviations from the simplified models. Both processes are capable of distorting or destroying the desired signal, and their noise-reduction effectiveness is dependent on the manner of application. Quality control methods are imperfect, and there is an element of human artistry in seismic data processing. Predictive deconvolution employs a one-dimensional (t) filter designed on a purely statistical basis. Seismic data represent observations in (x
1
, x
2
, t), a four-dimensional space, and multiple reflections are deterministically predictable (in principle) from the shallower primary reflection data. Common depth point stack operates in (x
1
, t) as an array steered for the arrival of the roughly spherical target reflection wavefront. Multidimensional filter design aimed at coherent noise rejection, based on the nonrandom noise structure in (x
1
, t) is often used in problem areas. More effective reduction of multiple reflections is one of many potentially useful improvements in the seismic method.