|
Seismic Attributes Consortium
Approach
At the present time Attrib3D generates attributes divided into 5
classes. The first two classes are instantaneous complex trace attributes,
which are computed sample by sample. The second two classes are
wavelet attributes. The first of these two classes is computed from
original trace data, and the second set is computed from signal-to-noise
improved traces. The fifth class is the geometrical attributes,
which are computed from the time/space relationship of the wavelets.
We will continue to define and classify new attributes based on
their utilization. In addition to the current list of 2-D and 3-D
poststack attributes we plan to develop new prestack attributes.
The Seismic Attribute Consortium has adopted the following approach:
-
Select seismic lines representing various
lithological settings and generate corresponding attributes
to investigate any visible indications
-
Select pre-processed seismic lines according
to the stratigraphic processing guide-lines, investigate effectiveness
of geometric attributes
-
Generate seismic attributes and analyze their
inter-relations by cross plots and cluster analyses
-
Generate seismic attributes and clusters for
specific attribute combinations by various clustering techniques
including the self-organizing maps
-
Generate seismic attributes and train Neural
Networks with interpreted well logs for predicting lithology
away from the training area
-
Develop pattern recognition techniques for
various stratigraphic settings
Current Deliverables
In Phases I and II of the consortium a number of sub-programs and
some interactive procedures have produced which can be used in either
a stand-alone mode or integrated within a more extensive analysis
and interpretation system. Following is a list of the products that
have been developed thus far:
-
Post stack attributes with physical characteristics,
such as amplitudes, frequency, velocity, Q, etc. (Note that
all routines are amenable to integration within different processing
systems)
-
Attributes with geometrical characteristics,
e.g. depositional bedding geometries such as parallel, diverging
or converging beds, chaotic patterns, event tracking, event
termination detection, termination modes, etc.
-
Interactive generation and display of 2-D
attributes
-
Attributes extended to 3-D
-
Interactive generation and display of 3-D
attributes
-
Associated n-dimensional clustering by Self-Organizing
Feature Maps and other clustering techniques
-
Provision of 4 output options for the computed
3-D attribute volume:
The current release of the Attributes Suite of
software is Attrib3D Version 2.6 with LithANN® beta 1.1 including
2-D and 3-D Attribute computation and display supported on SUNOS
5.5.1+ and IRIX 6.2 +
Applications of Research
The main emphasis is on the qualitative and quantitative measurement
of stratigraphic and lithologic effects on seismic signatures. These
measurements and their relationship to borehole measurements may
be used at a number of stages within seismic data processing and
interpretation sequences.
Stratigraphic data processing places significant emphasis on deriving
an accurate phase component and on maintaining a wide bandwidth.
Although important, amplitude and arrival times are not as critical
as the phase component. Stratigraphic interpretation begins with
an accurate picture of the depositional setting resolution, which
is principally controlled by the bandwidth and the phase. There
have been some papers published in IEEE journals claiming that useful
images can be obtained from the phase component alone and this needs
to be further investigated.
On the other hand the principal component in lithological processing
is firstly the amplitude information followed by the bandwidth.
For seismic sections used in structural interpretation, the overall
time/depth and velocity relationship becomes the most important
factor. It seems that while recognizing the importance of all of
these factors some of them become slightly more important depending
on the type of analysis conducted. Recognizing these facts, we can
intelligently design processing sequences that optimally lead to
structural, stratigraphic and lithologically viable seismic data
results.
Following are a few initial estimates of possible application areas.
We should note that when a tool is developed for a particular purpose,
we find it also to be useful in other areas as we gain experience
in its use.
1. Improvement of the stratigraphic processing
sequence. We will be able to establish more accurate processing
procedures that lead to more stratigraphically and lithologically
accurate seismic sections. We know that stratigraphic and lithological
objectives are not identical. However, for accurate interpretation
we need both to be accurately estimated. Additional processes
are projected to include surface-consistent amplitude, time and
phase statics computations, more consistent computation of Q factor,
wide-band zero phase section generation, and estimates of signal-to-noise
improved trace generation for AVO analyses.
2. Interpretation will be improved by the additional processes
and lithologic parameter estimates. Improvement will also
be effected by the development of a correct lithological processing
sequence. In this area we expect to generate additional seismic
attributes to reflect various physical parameters relating the
lithology of the subsurface. We expect to be able to generate
various synthetic logs to be derived from the seismic data or
from its attributes or from their combinations. It has been shown,
for example, that the combination of amplitude and phase provides
better continuity and therefore interpretability of seismic sections.
These routines will be developed so that they
can function as a part of, or can communicate with, a specifically
written interpretive program. Our goal is to prove that seismic
attributes contain information that can provide a more accurate
estimate of reservoir characteristics.
Current Work in Progress
-
Neural Network training with interpreted well
logs (under development as LithANN®)
-
Pre-stack attributes with physical characteristics,
e.g. most of the post stack attributes including those attributes
computed using trace offset as one of the variables such as
NMO or imaging, group and phase velocities, etc
-
Interactive event picking and picked event
attribute display and transfer and store for further computations
-
Interpreted well log or synthetic seismogram
overlay on attributes
-
Extrapolation of lithology away from wells
and display projected lithology
|
