Structural interpretation - 3-D Seismic Interpretation - page 62

passing through the earth; however, the effect of the earth filter on wavelet phase may
be hard to estimate. For detailed work involving measurement of loop amplitudes (see
chapter 5 ), it may be best to estimate a wavelet from the data. This is easily done by using
an algorithm that will calculate a wavelet that gives the best fit between the synthetic and
the real data. The goodness of fit can be evaluated from the cross-correlation coefficient
between the synthetic and the real seismic. However, the significance of high correlation
depends on the length of the wavelet compared with the analysis window. If the wavelet
is made long enough, then a perfect match can always be obtained, but such wavelets
are often implausible (e.g. having high-amplitude oscillations) because they are trying
to fit the noise in the real seismic as well as signal. A good match would be one with
a high correlation over a long gate using a short wavelet. An empirical approach is
to use a gate of 500 ms or more with a wavelet consisting of only 2-3 loops, but a
more rigorous approach has been put forward by White (1980) , who uses a statistical
method to constrain wavelet length. With this type of approach, not only is the wavelet
shape derived, but also the timing relative to time zero. This is important because
some minimum-phase wavelets can look approximately symmetrical, and so roughly
like a zero-phase wavelet, but the main loop is delayed from zero time. If we want to
measure amplitudes on seismic data, it is important to measure the right loop, e.g. the
one corresponding to the top of a reservoir.
In fig. 3.1 , a zero-phase wavelet of frequency content 5-50 Hz has been used. The
display repays close scrutiny. Firstly, it is a good idea to check the polarity of the display.
As mentioned in chapter 1 , there is often confusion about what the words 'normal' and
'reverse' mean as applied to the polarity of zero-phase seismic data. By looking at an
isolated interface with a sharp impedance change, the polarity of the synthetic can be
seen directly. Thus in fig. 3.1 there is a sharp increase in impedance at about 3020 ms
which corresponds to a white trough (deflection to the left) in the synthetic. Next, it
is easy to see where high-amplitude reflections are to be expected; these will be the
easiest events to pick on the 3-D dataset to form a basis for structural mapping. A sharp
impedance change gives the best response; the ramp-like impedance change at around
3200 ms causes only a moderate event, even though the total change in impedance is
large. The fine structure within the impedance log is not represented at all in the syn-
thetic. Detailed comparison for a particular target interval will show what hope there
is of mapping, for example, the top and base of a reservoir from the seismic data. It is
often useful to calculate synthetics for a range of high-frequency cutoffs, to see what
bandwidth would be needed to reveal significant detail; if a moderate increase in res-
olution would improve the information significantly, it is worth considering additional
seismic processing, inversion (chapter 6) , or re-shooting the survey.
The comparison of the synthetic seismogram with traces extracted from the
3-D dataset around the well location is shown in fig. 3.3 . In this case, there is a good
visual match for the main events, but not for the weak events in the central part of the
display. In such a case, it may be useful to calculate synthetic seismograms that include