The successful use of the technique we call ‘raypath interferometry’ has been documented in recent years on several 2D datasets. It has been applied not only in situations where near-surface corrections are large and non-stationary, but also where conventional statics solutions based on surface-consistency also work well. The raypath interferometry method is more complex than surface-consistent statics techniques, but because of the more general concept of ‘raypath consistency’, it is able to provide superior surface corrections for seismic data acquired in areas with complex surface conditions, or in situations like converted-wave (PS) imaging where near-surface variations can lead to large, non-stationary surface corrections.
To apply raypath interferometry to 2D data, the usual source (or receiver) gathers are transformed into a domain where raypath can be parametrized as a coordinate of the ensemble, using one of several possible transforms. This process is straightforward for 2D datasets; but extending the notion to 3D requires further consideration. We discuss here some possibilities and demonstrate one approach which seems promising. Introducing the surface azimuth between source and receiver locations, we demonstrate one pathway for extending raypath interferometry to 3D: extending the ‘surface function’ introduced in 2D raypath interferometry from a 2D function of surface location and raypath parameter to a 3D function of surface location, raypath parameter, and azimuth. We show one way of transforming 3D data into a useful raypath domain using the radial trace transform. We then demonstrate the creation of a 3D ‘reference wavefield’ and show that raypath interferometry, even for 3D data, can be implemented as a single cross-correlation and inverse-filter application, just as in the 2D case. Recovering the original 3D traces by reversing the data sorting/transform operation is straightforward. We show that various trace ensembles extracted from the corrected 3D data have no apparent residual statics compared to the same ensembles extracted from the raw data, hence confirming this approach.
View full article as PDF (5.44 Mb)