To summarize, field records contain (a) reflections, (b) coherent noise, and (c) random ambient noise. One important aspect of data processing is to uncover genuine reflections by suppressing noise of various types. Processing, however, cannot yield signal from field data without signal. At best, it suppresses whatever noise is in the field data and enhances the reflection energy that is buried in the noise. Seismic data must not be acquired with the attitude, “Don’t worry, processing will bring out the signal.”
Reflections on shot records are recognized by their hyperbolic traveltimes. If the reflecting interface is horizontal, then the apex of the reflection hyperbola is situated at zero offset. On the other hand, if it is a dipping interface, then the reflection hyperbola is skewed in the updip direction.
There are several wave types under the coherent noise category:
- Ground roll is recognized by low frequency, strong amplitude, and low group velocity. It is the vertical component of dispersive surface waves. In the field, receiver arrays are used to eliminate ground roll. Ground roll can have strong backscattered components because of lateral inhomogeneities in the near-surface layer.
- Guided waves are persistent, especially in shallow marine records in areas with hard water bottom. The water layer makes a strong velocity contrast with the substratum, which causes most of the energy to be trapped within and guided laterally through the water layer. The dispersive nature of these waves makes them easy to recognize on shot records. Guided waves also make up the early arrivals. The stronger the velocity contrast between the water layer and the substratum, the smaller the critical angle; thus, more guided-wave energy is trapped in the supercritical region. When there is a strong velocity contrast, refraction energy propagates in the form of a head wave. Guided waves also are found in land records. These waves are largely attenuated by CMP stacking. Because of their prominently linear moveout, in principle they also can be suppressed by dip filtering techniques. One such filtering technique is based on 2-D Fourier transformation of the shot record. This is discussed in frequency-wavenumber filtering. Another approach is based on slant stacking, which is described in the slant-stack transform.
- Side-scattered noise commonly occurs at the water bottom, where there is no flat, smooth topography. Irregularities of varying size act as point scatterers, which cause diffraction arrivals with table-top trajectories. They can be on or off the vertical plane of the recording cable. These arrivals typically exhibit a large range of moveouts, depending on the spatial position of the scatterers in the subsurface.
- Cable noise is linear and low in amplitude and frequency. It primarily appears on shot records as late arrivals.
- The air wave with a 300 m/s velocity can be a serious problem when shooting with surface charges such as Geoflex, Poulter, or land air gun. Perhaps the only effective way to remove air waves is to zero out the data on shot gathers along a narrow corridor containing this energy (notch muting). It often is impossible to recover any data arriving after the air wave on Poulter data.
- Power lines also cause noisy traces in the form of a monofrequency wave. A monofrequency wave may be 50 or 60 Hz, depending on where the field survey was conducted. Notch filters often are used in the field to suppress such energy.
- Multiples are secondary reflections with interbed or intrabed raypaths. Guided waves include supercritical multiple energy. Multiples are attacked by methods, which are based on moveout discrimination, and prediction theory, which uses the periodic behavior of multiples. The most effective moveout-based suppression technique often is CMP stack with inside-trace mute (multiple attenuation in the CMP domain). Prediction theory should be particularly effective, at least in theory, in the slant-stack domain.
Random noise has various sources. A poorly planted geophone, wind motion, transient movements in the vicinity of the recording cable, wave motion in the water that causes the cable to vibrate, and electrical noise from the recording instrument all can cause ambient noise. The net result of scattered noise from many scatterers in the subsurface also contributes to random noise .
In gain applications, it is noted that energy propagating within the earth is subject to a decay in amplitude because of wavefront divergence and frequency-dependent absorption from the intrinsic attenuation of rocks. Signal strength therefore decreases in time, while random noise persists and eventually dominates. Unfortunately, gain corrections to restore signal strength at later times boost random noise in the process. Fortunately, CMP stack suppresses a significant part of the random noise uncorrelated from trace to trace.
- Larner et al., 1983, Larner, K. L., Chambers, R., Yang, M., Lynn, W., and Wai, W., 1983, Coherent noise in marine seismic data: Geophysics, 48, 854–886.
- Introduction to fundamentals of signal processing
- The 1-D Fourier transform
- The 2-D Fourier transform
- Gain applications
- Basic data processing sequence