Introduction

The usual strategy for the design of 2-D and 3-D seismic surveys starts from the estimation of ``critical'' values for subsurface parameters such as maximum geological dip, maximum target depth, minimum target rms velocity and minimum target thickness. From these values (and an estimation of the minimum required fold), we choose acquisition parameters such as receiver group and source interval, maximum offset and number of active channels per shot. In 3-D we also compute the number of active receiver lines, the number of shots per salvo, the width of the active patch and the geometry of the acquisition template (). More often than not these parameters are held constant across the whole survey irrespective of changes in the geometry of the subsurface we wish to image or of its associated velocity field.

Common usage in 3-D seismic acquisition focuses on regularity of source and receiver positions on the surface of the earth according to one of a few ``standard'' geometries. These geometries are designed to provide as regular offset and azimuth coverage in each CMP bin as possible while at the same time allowing for relatively easy logistics. Although there are important differences between the different ``standard'' geometries in terms of their offset and azimuth distributions () and other less obvious characteristics such as the symmetric sampling of the wavefield (), these regular geometries all share these characteristics:

• Quality control is essentially qualitative and relies on regularity of offsets and azimuth in adjacent CMP surface bins. The actual subsurface bin population at the target location (which is clearly a function of the reflector dip and the dips and velocities of the overburden) is completely ignored or considered only as an afterthought of the design.
• Worst of all, these geometries make no attempt whatsoever to guarantee or even promise any sort of uniformity in subsurface illumination. This may be critical when trying to image under salt flanks and in overthrust areas with large fault displacements (which are common in foothill data). No amount of clever processing can compensate for problems generated by poor illumination during acquisition.

The choice of the parameters for these ``standard'' geometries can be posed as an optimization problem once the basic template is chosen () and even the logistics and economics of the acquisition can be incorporated into the computation, at least to some extent (). The basic premise, however, is that the underlying acquisition template is regular and chosen before hand, so that the optimization is restricted to look for the best combination of the parameters consistent with the chosen template.

The quality of the overall design is evaluated based on such surface attributes as uniformity of fold of coverage and regularity of offset sampling (and azimuth sampling in 3-D). This is equivalent to an implied assumption of a layered, constant velocity subsurface model. Illumination is only considered (if considered at all) as a forward problem. Illumination maps may be constructed at the target reflectors for each of a few competing geometry templates and the best one selected by a qualitative comparison of those maps (). Moreover, the complete design may be simulated in the computer further analyzing the resultant distribution of offsets, azimuths and intensity of illumination (). Although this is an important step in the right direction, there is no guarantee that the chosen geometry will in fact produce optimum illumination.

I propose to base the design on an initial structural and velocity model of the subsurface, even if only a crude one. Obviously, at the time of acquisition, we don't have a detailed subsurface model. Oftentimes, however, we know the rough features we wish to image. Is it a salt dome, an overthrust faulted anticline or a deep channel turbidite system? This information may come from previous seismic data, from well logs, from a conceptual geological model, from surface geology and usually from a combination of all of them. This wealth of information is ignored in the usual acquisition design but it does not need to be. It is possible and indeed desirable to use the existing knowledge of the subsurface structure and velocity model to improve the acquisition of new data. I show with a very simple synthetic 2-D example that we can reduce the number of shots without compromising the quality of the image by selectively ignoring shots whose contribution to the image is less than, say, half the number of traces of a regular shot.

The two key points are: we are not required to use the same parameters or indeed the same geometry all across the survey and regularity of surface parameters is not necessarily the mark of an optimum design. A better indicative is regularity of subsurface attributes such as target illumination.