Validating vertical velocity gradients in near-surface refraction seismology

Abstract

Wavepath eikonal traveltime (WET) refraction tomograms are generated with the generalised reciprocal method (GRM), a novel medium resolution common offset gather (COG) implementation of the GRM, using uniform velocities and vertical velocity gradients, and the low resolution default starting model consisting of smooth vertical velocity gradients. All tomograms have comparable misfit errors, which illustrates the ubiquity of non-uniqueness and the necessity for validating all starting models. Nevertheless, the use of even the maximum vertical velocity gradients in the weathered region does not produce any improvement in the spatial resolution of the seismic velocities in the sub-weathered region with either the default or the COG GRM starting models. Therefore, if a low resolution starting model is used, then the most likely outcome will be a low resolution WET tomogram, irrespective of whether or not vertical velocity gradients are employed. Vertical velocity gradients can be represented as part of a continuum of seismic velocities in the weathered layer, which range from uniform to hyperbolic velocities, and which are consistent with the traveltime data. Acceptable models employ seismic velocities in the weathered and sub-weathered regions computed with the same XY value. In this study, the optimum XY value is representative of uniform seismic velocities, whereas the maximum XY value, which is the average cross-over distance, is representative of default and hyperbolic velocities. Intermediate XY values indicate more moderate vertical velocity gradients and/or undetected layers. WET tomography is largely a smoothing operation which does not improve the spatial resolution of either medium resolution or detailed starting models. A limited number of iterations, usually a maximum of five, can improve the cosmetics of refraction tomograms, whereas repeated applications can result in considerable reductions in spatial resolution, and eventually, low resolution tomograms consisting of essentially featureless parallel quasi horizontal layers.