Académique Documents
Professionnel Documents
Culture Documents
October 2009
INTERPRETERS CORNER
is, with the new criterion, h/4 is not
necessarily the minimum thickness of
reservoir units for stratigraphic or owunit correlation.
90- or zero-phase wavelet?
Use of a 90-phase wavelet in expanding Rayleighs criterion (Figure 2) leads
to an interpreter-friendly composite
waveform for BR. This observation
is based on the fact that a composite
waveform from one thin bed resembles
the derivative, or 90-phase equivalent,
of the input wavelet. As a result, the interference pattern for a two-bed model
has a waveform similar to that of the
two-interface model in Rayleighs criterion (Figure 1).
For comparison, if a zero-phase
Ricker wavelet is applied to the two-bed
model in Figure 2, the synthetic trace
demonstrates a far more complex interference pattern (Figure 3). Interpreters
must be able to recognize a peak-trough
couplet for each bed for fair resolution
of the thin beds. At the resolution limit
achieved using the 90-phase Ricker
wavelet (Figure 2e), only one peak and
one trough are apparent, and the thin
beds cannot be considered resolved
(Figure 3e). Consequently, a zero-phase
wavelet has less thin-bed resolving
power than its 90-phase equivalent.
The resolution limit (BR) for the zerophase wavelet cannot be dened until a
thicker barrier (b2) is inserted (Figure Figure 4. Idealized amplitude versus frequency (AVF) plot for a randomly generated thin-bed
AI prole (yellow line). Seismic traces are results of the RC series of the AI prole convolving
3f ). Besides, in practice, to carry out with 90-phase Ricker wavelets of varying predominant frequencies. In the thin-bed range,
correlation of thin beds by tracking one sandstones tie to trough (red) events. BR=bed resolution picked at the point where the sandstone
peak-trough couplet for each bed is con- event is about to merge with other events; IR=interface resolution interpreted at the tuning
siderably more dicult than the simple point with maximum amplitude.
peak-to-peak (or trough-to-trough, depending on AI prole) tracking required for thin-bed correla- mic display for this purpose is called amplitude versus frequency (AVF). AVF was rst proposed for studying dependence of
tion in 90-phase data (Zeng and Backus, 2005).
seismic facies on data frequency (Zeng et al., 2000). An ideal
Amplitude versus frequency analysis
AVF prole would be one illustrating seismic responses of a
Selection of tools for resolution analysis depends on the pur- given AI prole to a group of wavelets that are characterized
poses of a study. To test the resolution power of a wavelet, by the same (or at least similar) phase and band ratio but
seismic modeling of an interbedded wedge model that al- vary in predominant frequencies. In a synthetic example (Figlows the thickness of one or more beds to vary with distance ure 4), a reection coecient series of a random, multiple,
would be ideal. Results in Figures 13 are from convolution thin-bed prole is convolved with a series of 90-phase Ricker
models using this method. Noise and modeling/processing wavelets whose predominant frequencies range from 8 to 80
Hz. In this more realistic model, it is observed that, when
eects can be added if desired.
If the goal is to determine resolution limits (BR and IR) frequency is high, all thin beds are resolved in IR. When freof individual reservoir units in a geologic section, either in quency decreases, pairs of the closest two thin beds start to
synthetic or eld-data cases, the wedge model-based approach merge (BR); then merged ones continue to interfere with one
is inecient, and a site-specic analysis of amplitude interfer- another, leading to more merging, and so forth. For any bed of
ence patterns in time versus frequency is necessary. The seis- interest (a sandstone in Figure 4), the predominant frequency
October 2009
1193
INTERPRETERS CORNER
at the resolution limit (BR frequency)
can be determined at the event merging
point if the seismic correlation is known
(negative or red events for sandstones in
Figure 4), which then can be converted
to a resolution limit in time or wavelength. Similarly, one can follow the
tuning trend for each event and interpret tuning frequency at the maximum
amplitude, which corresponds to IR
(Figure 4). BR and IR frequencies can
thus be picked for all thin beds where
resolution allows.
In practice, one has to work with
eld-data traces at well sites. With only
one trace for a well prole, frequencydriven signal decomposition must be
performed on the trace to nd the AVF
relationship. Relevant methods include
digital ltering (e.g., FFT), wavelet
transform, spectral decomposition, etc.
The challenge is to achieve enough temporal and frequency resolution without
signicantly distorting the interference
patterns. It is currently a subject of
research. A test using the digital ltering method produces an AVF plot of
a well-site eld trace from the Miocene, oshore Louisiana, Starfak Field
(Figure 5). Although the eld data are
not noise-free and wireline logs reveal
more complex bedding architecture, including transitional boundaries, generated AVF shows reasonable amplitudefrequency trends for most sandstones
that can be used to pick BR- and IRfrequency points.
Figure 5. Field seismic section and wireline logs (GR/RES), well-site seismic trace, and AVF in
Miocene Starfak Field, oshore Louisiana. The seismic trace is 90 phased, with troughs (red)
indicating low-AI sandstones. Seismic data are relatively low in predominant frequency (30
Hz in shallow and 20 Hz in deep sections). The nonlinear frequency scale is due to nonlinear
conversion from the lter panel to predominant frequency values. Arrow indicates tuning trend
for Sequence 2 sandstone.
October 2009
INTERPRETERS CORNER
1195
INTERPRETERS CORNER
October 2009
INTERPRETERS CORNER
hopefully in turn improve many aspects of geophysics, from
exploration and reservoir characterization to eld surveillance. Workows that integrate analyses of both IR and BR
should also improve reservoir prediction and correlation. Immediate applications include, but are not limited to:
Predicting reservoir thickness thinner than h/4 using BR
with a single seismic trace,
Improving reservoir thickness mapping up to h/4 using IR
with a single seismic trace,
Amplitude detuning of thin reservoirs for DHI with a
single seismic trace,
Optimal facies imaging by adjusting
predominant frequency of seismic facies to a thickness-dependent tuning
frequency,
Frequency-dependent high-resolution
seismic stratigraphy.
Conclusions
Better seismic resolution can be achieved
by shifting the focus of seismic interpretation from top and bottom interfaces
of a thin bed to the thin bed itself. Th is
task can be accomplished by rst dening bed resolution using an expanded
Rayleighs criterion and then conducting AVF analysis on the model and
eld-data examples. Bed resolution has
no limit for noise-free data and can be
as small as h/16 for eld data using current technology. For maximum benet,
seismic data must be adjusted to the 90
phase.
Suggested reading. For Rayleighs criterion please refer to Fundamentals of
Optics by Jenkins and White (McGraw
Hill, 1957). The classic discussion of the
seismic resolution limit was by Widess
in How thin is a thin bed? (Geophysics, 1973) and summarized in The limits
of resolution of zero-phase wavelets by
Kallweit and Wood (Geophysics, 1982).
For benets of 90-phase data in thin-bed
interpretation, read Interpretive advantages of 90-phase wavelets, Part I: Modeling and Part II: Seismic applications by
Zeng and Backus (Geophysics, 2005).
SEG published the fourth edition of Sheris Encyclopedic Dictionary of Applied
Geophysics in 2002. Amplitude versus
frequencyapplications to seismic stratigraphy and reservoir characterization)
October 2009
1197