THE HISTORICAL DEVELOPMENT OF DEEP REFLECTION SEISMIC DATA ACQUISITION AND PROCESSING

Although the most powerful geophysical technique, deep reflection seismic profiling has usually been undertaken at very regional scales. Processing schemes applied to such data are conventionally simplistic, mainly a symptom of the problems in differentiating between (weak) signal and noise. The long-standing attitude has been that complex processing routines can not be used with confidence or reliability. Many of these problems of course relate to the unique characteristics of deep seismic data. Migration has been perhaps the biggest challenge, due to the large aperture inherent in a 12-20 s TWT section (Warner, 1987). Even the fundamental aspects of the seismic method have associated complexities. Of interest to any seismic study, the accurate determination of rock velocities are of fundamental importance. From a regional standpoint, wide-angle refraction seismic profiling is often preferable to reflection seismic profiling, as rock velocities can be better estimated using refraction data. Advanced first-arrival inversion routines (e.g. Hole, 1992) are now able to produce detailed 2.5D tomographic velocity models, even when the acquisition involved irregular and crooked recording geometries. In contrast, the lack of perceptible moveout at depth on CDP gathers makes velocity determination difficult and imprecise. It is not uncommon that refraction data are used to constrain the velocity models input into CDP stacking of reflection seismic data.

Crustal-scale surveys are of course acquired over very large distances, commonly several hundred kilometers, which will usually involve traversing rugged and variable topography. Often these problems are so extreme that remote recording stations are left switched on in (passive) recording mode for several days, while various shots are detonated along the line. The continuous records are then deciphered back at home. This style of recording dictates a coarse receiver station spacing, which in turn dictates that the data will be dominated by long wavelengths and low frequencies. As such, the required processing scheme is relatively simplistic by comparison to that required for imaging of shallow exploration data.

Furthermore, most processing schemes are unique to the particular datasets being processed. An example is the imaging of the major features beneath mid-ocean spreading ridges, where new oceanic crust is formed and rafted away on diverging tectonic plates. Figure shows the oceanic crust to be composed of 3 (seismically) distinct layers: layer 2a (sediments, pillow lavas and extrusives) which varies between 200-500 m thick, layer 2b (sheeted dykes) which is 2-3 km thick, and layer 3 (gabbros) which extends down to the common oceanic crustal depth of 7-8 km. Beneath the mid-ocean ridge (MOR), an upwelling axial magma chamber (AMC) supplies the material required to generate new crust as the plates diverge. The two features commonly of most interest are the variable thickness of the 2a layer (which reveals much about the tectonic evolution of the upper-crust) and the width of the AMC (which reflects the ``magma budget'', a key control upon the along-axis structure and morphology of the ridge).

 

fig1
Figure 1 Schematic drawing of the East Pacific Rise mid-ocean ridge. The faulted seafloor forms the top of layer 2a (sediments and extrusives). Layer 2b is entirely composed of sheeted dikes, while layer 3 is predominantly gabbroic. A magma chamber provides the material emplaced between the diverging plates. From Kent et al. (1993).

 

fig2
Figure 2 Seismic profile from the East Pacific Rise. The most obvious features resolved after stacking three times with different offset ranges and then merging, are the base of layer 2a (far offsets), the seafloor and top of the AMC (near offsets). Conventional processing was unable to properly image the various features.

As a pronounced velocity gradient exists within the 2a layer (from 1.5 - 5 km/s), many of the recorded seismic events are from turning ray energy, which is commonly only recorded at offsets greater than 1.5-2 km. Conversely, the best resolution of both the seafloor and the AMC are obtained from near-vertical reflections. Hence, one processing approach successfully used has been to produce 3 separate stacks; 2 using near-offset reflections to image the seafloor and AMC respectively, and 1 using far-offset reflections to image the base of the 2a layer. These 3 stacks are then merged to produce a single output stack (Fig. ). Such an approach does not favor the subsequent successful application of migration, a problem further compounded by the presence of significant diffraction energy related to seafloor topography, which is often out-of-the-plane.

This is of course only a single, isolated example. Nevertheless, it demonstrates the often unique and simplistic approaches required to image regional deep seismic data.