SUMMARY A reliable and clear depth image requires accurate near-surface velocity information. Such valuable data are usually expensive to obtain from an uphole program. Here, we show that a fairly accurate model of the shallow subsurface can be obtained from high-resolution seismic data. Using conventional velocity analysis and tomographic inversion, we build a near-surface model for a real-data example. The nearsurface model results in a better near-surface correction than what is obtained using conventional refraction-based static methods. The accuracy of the model is confirmed by a comparison with uphole data and the stack quality using the model.
INTRODUCTION The problem of the complexities in the near-surface and their influence on seismic imaging have been covered extensively in the literature. The main conclusion is that an accurate velocity model is required to overcome this problem. In his book, Cox (1999) provides a comprehensive account of the problem and the existing methods which are used to build a near-surface velocity model. The two most widely used classes of techniques are the uphole-based and the refraction-based methods. The problems associated with existing methods were our motivation to propose an alternative or a complementary technique. In an earlier work (AlZayer and AlKhalifah, 2006), we proposed using shallow high-resolution seismic data to construct the near-surface velocity model. In a subsequent work, we demonstrated a technique to make such data more cost-effective. Recording reflected and/or refracted waves from shallow interfaces is not as easy as recording conventional seismic. The shallow-seismic technique, however, is more developed and widely used today than anytime (Steeples et. al, 1990). Refracted waves are probably easier to record and identify since they suffer less from interference with other modes, such as surface waves, but require larger offsets to develop. Once we record reflection or refraction arrivals, our next challenge is to analyze the data to gain desired Earth parameters, such as velocity, density, or some engineering parameters. In this abstract, we investigate different methods of obtaining velocity information from reflection and/or refraction time-arrivals. There exit a number of methods to obtain velocity from seismic data. They, however, can be classified into two classes: forward and inverse methods. In the forward class, a suite of functions are used to correct the data. The best-result functions are picked at selected locations, which are then interpolated to build the model. Examples of such methods are the conventional coherency analysis and some of the migration velocity-analysis techniques. Under the inverse class, picked functions, such as travel-time or dispersion curves, are inverted using a mathematical model. Of course, the accuracy of any method depends on the accuracy of the picks as well as the mathematical model. Reflection and refraction tomography are classical examples of such techniques. Some algorithms use a combination of the two approaches; a forward technique is used first to obtain an initial model. The residual errors are then inverted to refine the velocity functions.
Coherency Analysis The most direct way of velocity estimation is the conventional hyperbolic, or non-hyperbolic, semblance velocity analysis.