Coherent linear noise
Two types of coherent linear noise that deserve special attention are guided waves and side-scattered energy. Figure 6.0-1 shows field data with coherent linear noise in three different domains — common-shot gather, common-midpoint gather, and CMP stack. The dispersive waves labeled as A in Figure 6.0-1a are guided waves, and the linear events B and C and the events D with curvature are associated with side scatterers. Guided waves manifest themselves as dispersive linear noise on both common-shot and CMP gathers, but are attenuated largely by stacking.
Guided waves are trapped in a water layer or in a low-velocity near-surface layer and travel in the horizontal direction. They are dispersive — each frequency component propagates with a different phase velocity, and are best described by normal-mode propagation. Numerical modeling of guided waves using normal-mode propagation theory is presented in Section F.1. Since they do not contain any useful reflection energy, guided waves usually are muted on CMP gathers. When one mode splits away from the rest of the guided wave packet and travels at lower speeds, and thus overlaps with reflection events, then dip filtering in the f − k domain is needed (frequency-wavenumber filtering).
An excellent example of guided waves is seen on the field record (between 1 and 4.5 s at far offset) shown in Figure 6.0-2. The first part of the wave packet between 1 and 1.5 s at far offset contains low frequencies. High frequencies ride along the direct arrival path at approximately 0.3 s at near offset and 1.8 s at far offset, followed by moderate frequencies between 1.8-2.8 s at far offset. The very-low frequency dispersed wavetrain with high amplitudes between 2.8-4.5 s at far offset corresponds to the mode that has split from the rest of the guided-wave modes. This happens when there is a shallow, soft water bottom associated with a mud layer. Also seen on this record is the backscattered guided wave (zone B) with reverse linear moveout, which indicates the presence of irregularities over the ocean bottom. These irregularities also cause arrivals (zone A) that represent point scatterers.
The dispersive nature of guided waves can vary along a seismic traverse depending on water depth and water-bottom conditions. The shallower the water depth and the softer the water bottom, the more the dispersion and splitting of modes associated with guided waves (Figure 6.0-3).
Side-scattered energy has a large moveout range depending on the position of the scatterer acting as a point source at the water bottom with respect to the position of the recording cable (events B, C, and D in Figure 6.0-1a). Side-scattered energy manifests itself with varying moveout on common-shot gathers (Figure 6.0-1a), and is not apparent on CMP gathers (Figure 6.0-1b), but reappears as linear noise on stacked sections (Figure 6.0-1c) (Larner et al., 1983).
Side-scattered energy stacks at high velocities along the linear flanks of its traveltime curve. We then anticipate that the linear noise seen on a stacked section, particularly at late times, most likely is scattered energy along the flanks of its traveltime curve, stacked together with high-velocity primary energy (Figure 6.0-4).
Linear noise associated with side scatterers is recognized easily on time slices from a 3-D volume of stacked data. Note in Figure 6.0-5 the circular patterns expanding out from the source of a series of point scatterers at the water bottom. In this case, certain parts of the sea-bottom pipelines act as point scatterers.
Attenuation of coherent linear noise associated with side scatterers may be achieved by f − k filtering (frequency-wavenumber filtering), τ − p transform (the slant-stack transform) or Radon transform (the radon transform) techniques. A linear event on a shot record maps onto a radial line in the f − k domain, and thus can be rejected by f − k dip filtering. A linear event on a shot record maps onto a point in the τ − p domain, and thus can be rejected by muting in the τ − p domain. Finally, spatially random noise and coherent linear noise are not included in the mapping from CMP domain to Radon-transform domain based on hyperbolic moveout. As a result, the reconstructed CMP gather by way of inverse transform will be free of noise.
Coherent linear noise also exists in land data in the form of dispersive Rayleigh wave, commonly known as ground roll. This type of coherent noise has low group velocity and large amplitudes and is limited to low frequencies. In fact, as shown in Figure 6.0-6, ground-roll energy almost always dominates the reflection energy that may be present in the recorded data. Only after some type of amplitude scaling, reflections become visible (Figure 6.0-7). Note from the selected shot records in Figure 6.0-6 that the dispersive waves associated with the ground-roll energy change in strength and stepout (the dip of the linear noise trend) as a result of variations in the near-surface conditions.
Swell noise manifests itself on shot records in the form of low-frequency vertical streaks (Figure 6.0-8). This type of noise arises from rough weather conditions during marine seismic recordings, especially in shallow waters. A low-cut filter often removes the swell noise from shot records.
Finally, cable noise is one other type of coherent noise that manifests itself in the form of low-frequency linear events with very large stepout as seen on the shot records in Figure 6.0-9. Note the increase in the energy level of the cable noise as the water depth becomes shallower. As for the swell noise, a low-cut filter often removes the cable noise from shot records.
Figure 6.0-5 A time slice from an unmigrated 3-D volume of stacked data which exhibits circular patterns associated with point scatterers along sea-bottom pipelines. (Data courtesy Total Argentina.)
- Treatment of coherent linear noise by conventional processing
- Reverberations and multiples
- Treatment of reverberations and multiples by conventional processing
- Spatially random noise