Geology Reference
In-Depth Information
Fig. 4.19
Filter panels showing the frequency content of a panel of reflection records by passing them through a series of narrow-band
frequencies.This plot allows the geophysicist to assess the frequency band that maximises the signal-to-noise ratio. Note that this may vary
down the traces due to frequency-dependent absorption. (From Hatton
et al
. 1986, p. 88)
main types of waveform manipulation are frequency fil-
tering and inverse filtering (deconvolution). Frequency
filtering can improve the SNR but potentially damages
the vertical resolution, while deconvolution improves
the resolution, but at the expense of a decrease in the
SNR. As with many aspects of seismic processing, com-
promises must be struck in each process to produce the
optimum overall result.
Since the dominant frequency of reflected arrivals
decreases with increasing length of travel path, due to the
selective absorption of the higher frequencies, the char-
acteristics of frequency filters are normally varied as a
function of reflection time. For example, the first second
of a 3 s seismic trace might typically be band-pass filtered
between limits of 15 and 75 Hz, whereas the frequency
limits for the third second might be 10 and 45 Hz. The
choice of frequency bands is made by inspection of filter
panels (Fig. 4.19). As the frequency characteristics of
reflected arrivals are also influenced by the prevailing
geology, the appropriate time-variant frequency filter-
ing may also vary as a function of distance along a seismic
profile. The filtering may be carried out by computer
in the time domain or the frequency domain (see
Chapter 2).
4.8.1 Frequency filtering
Any coherent or incoherent noise event whose domi-
nant frequency is different from that of reflected arrivals
may be suppressed by frequency filtering (see Chapter
2). Thus, for example, ground roll in land surveys and
several types of ship-generated noise in marine seismic
surveying can often be significantly attenuated by low-
cut filtering. Similarly, wind noise may be reduced by
high-cut filtering. Frequency filtering may be carried
out at several stages in the processing sequence. Nor-
mally, shot records would be filtered at a very early stage
in the processing to remove obvious noise. Later applica-
tions of filters are used to remove artefacts produced by
other processing stages. The final application of filters is
to produce the sections to be used by the seismic inter-
preters, and here the choice of filters is made to produce
the optimum visual display.
4.8.2 Inverse filtering (deconvolution)
Many components of seismic noise lie within the fre-
quency spectrum of a reflected pulse and therefore
cannot be removed by frequency filtering. Inverse filters
discriminate against noise and improve signal character
using criteria other than simply frequency.They are thus
able to suppress types of noise that have the same
frequency characteristics as the reflected signal. A wide
range of inverse filters is available for reflection data
