Geology Reference
In-Depth Information
indicated by curved, upwardly convex
diffraction patterns
createdbyaform
of sideswipe. The root cause of the problem is that A-scans are plotted out
vertically to form B-scans, and therefore all events, even if from reflection
along a slant path, appear vertically below the surface antenna positions.
Small reflectors or sharp angles in continuous reflectors can reflect radar
waves at many angles, producing diffraction patterns (Figure 10.20a). Pro-
vided that velocities vary only with depth, these patterns are hyperbolic. The
source, provided that the traverse line actually passes over it, is located at
the pattern apex.
Less dramatic distortions affect the positions of dipping reflectors, since
reflections from these also travel along slant paths (Figure 10.20b). Tight,
upwardly concave
synclines
can produce three reflections at a single surface
point, giving rise to the peculiar features known as
bow-ties
(Figure 10.20c).
All these distortions can be corrected by the
migration
programs first
developed for use with seismic data, but these must work on numerous
traces simultaneously and cannot easily be applied in the field.
It is almost true to say that radar applications are divided into two distinct
groups. In the first, layered structures are imaged and mapped. In the second,
diffraction patterns are sought that reveal the presence of limited highly
reflective targets.
10.3.3 The processing sequence
As already noted, most of the processing techniques now being used for
GPR data resemble those developed for seismic data, and seismic processing
software has been used almost unmodified to enhance GPR results. There
are differences in emphasis, largely because of the well controlled nature of
the radar pulse and the general use of single-fold rather than CMP coverage,
but these need not concern the operator in the field.
After stacking, GPR data are passed through low-cut filters, to remove
noise due to inductive effects and the limitations in instrument frequency
response, and high-cut filters, to eliminate noise spikes. The decrease in
signal amplitude with time is then reversed by time-variant amplification.
Automatic gain control
(AGC) is used to do this in the field, producing
records for quality control, but data are normally stored in unmodified form.
Compensation, in the processing centre, for propagation effects, using SEC
(
spherical and exponential compensation
) filters based on physical models
of the subsurface, comes after any frequency-based filtering, because time-
variant gain functions distort wavelets and must be applied with care if
amplitude integrity is to be preserved. Migration algorithms operate on the
entire data set rather than individual traces (Figure 10.21), and are usually
the last to be applied.
