"DIAGNOSTIC 3-D SEISMIC PROCESS”
D3DSP* DATA PROCESSING GUIDE
L. Willhoit 4/12/2002 * Patent #5,671,136
GOM EXAMPLE DESCRIPTION (To be modified, as needed):
1. INITIAL PRE-PROCESSING
Processing begins with Field Recorded
Seismic Data tapes. The data usually has
no geometry assigned
to the trace
headers. Data traces will be converted
to 2ms sample interval, using an “intelligent” sample
interpolating
algorithm. Obviously bad traces should
be identified and deleted here.
2. COMPENSATION FOR INSTRUMENT IMPULSE RESPONSE
& GEOMETRIC SPREADING
To remove the effects of the recording
instrument an inverse filter will be constructed to convert the
instrument
response to a zero-lag spike. This
filter will be applied through convolution with the seismic
data. As a correction for spherical divergence, a
time-variant geometric spreading scale function will be
applied to the
entire volume. Shot records will be
displayed before and after application of the
geometric-spreading
function for QC purposes.
3. TIME VARYING (TIME-DOMAIN) SPECTRAL BALANCING
(TVSB)
Time-varying spectral balancing will be
applied to the volume to whiten the spectrum in the (GOM)
6-125 Hz. frequency
range. Minimum-phase bandpass
filters are recommended for spectral
decomposition. Shot records and amplitude spectra will be
displayed pre- and post-TVSB.
4. TOMOGRAPHIC / REFRACTION STATICS
Apply if available. Not available (or required?) in many offshore
areas.
5. SURFACE-CONSISTENT DECONVOLUTION
Usually a three-step process involving
spectral analysis, surface-consistent frequency decomposition, and
operator
design/application. The data may be
scaled with a 500 ms AGC gain to balance amplitudes
prior to spectral
analysis. The AGC gain is not applied to
the actual data to which the decon operators
are to be
applied. Multiple hyperbolic windows may
be used for spectral analysis, having varying time
extents at zero
offset.
The prediction distance will be one sample
long. The operator length is to be
determined from test
Autocorrelograms
using lengths of 40, 64, 128, and 256 ms, with 0.1% white noise. The spectra will be
decomposed into
source and detector terms prior to operator design and application.
6. AMPLITUDE ANOMALY PROCESSING (AAP)
Attenuates anomalous
amplitudes in a surface consistent manner. RMS amplitudes can be measured over
500 ms. windows of all
data. Amplitude statistics can be
decomposed into source, detector, offset and
CMP terms. The decomposed amplitudes can then be
compared to the RMS amplitudes of the individual
traces and
anomalous amplitudes on each trace can be attenuated. Amplitude analysis should be
performed on a
swath-by-swath basis, but amplitude decomposition should be run on all data
simultaniously.
7. SURFACE-CONSISTENT AMPLITUDE COMPENSATION
Need to produce surface-consistent shot and
detector multipliers for the entire survey, analogous to
Residual Static
corrections.
8. RESIDUAL AMPLITUDE COMPENSATION (OFFSET ONLY)
An offset-consistent scaling routine will
be used, which compensates for the higher amplitudes
associated with
the near-offset traces. Time-variant
amplitude analysis should be performed on a subset
of shots from the
entire survey. These shots will have
Amplitude Anomaly Processing and Surface
D3DSP Processing Guide Page
2 of 3 4/11/2002
Consistent Amplitude Compensation applied,
as well as the time-variant filter used in SCAC picking.
This scaling routine should separate the
amplitude statistics into offset groupings (roughly twice the
receiver spacing)
and derive a set of time-variant offset-dependent multipliers for each
group. These
scalars should
then be applied to the Anomaly/SC Amplitude corrected, unfiltered volume
according to
the offset of each
trace.
9. AMPLITUDE ANOMALY PROCESSING (CMP ONLY)
A final pass of AAP (Par. 6), above, should be run
decomposing the CMP term only. With
shot,
detector and
offset balancing complete, any remaining amplitude anomalies should reside in
the CMP
domain. The same windowing parameters used in the
earlier AAP run will be incorporated in this pass.
10. PRELIMINARY VELOCITY ANALYSIS
Velocity analyses will be derived at
roughly a 5000' interval in the inline direction and a 4000' interval in
the crossline
direction. At each location along an
inline, five-to-seven CMP's will be summed for the
velocity spectra.
Multi-velocity function stacks and gathers will also be used as velocity
interpretation
tools. Velocity picking will be performed in an
interactive velocity processing system, in which the
velocity picks are
used to interactively NMO-correct the associated CMP gather. By doing this, the
accuracy of the
moveout correction can be seen immediately.
The picks will also be overlaid on the
corresponding
multi-function stack, so that the stack response can be observed. Preliminary velocity
analysis should be
performed on deconvolved data with the refraction or tomographic refraction
solution
(if any)
applied. The data input to velocity
analysis can also be filtered with a 10-45 Hz, zero-phase
filter, and scaled
with an RMS gain.
11. REFLECTION RESIDUAL STATICS (1ST PASS)
A static-time-correction routine is
required that resolves observed reflection times into surface-consistent
residual source
and detector static corrections. Using a
Gauss-Seidel-type method, a least-squares error
solution for the
reflection statics problem should be abtained. For the first pass, static time deviations
should be picked
against a stacked model dataset in a “good data” window, with a reasonable
(24ms?)
time shift
limit. The data input to picking should
be nmo-corrected decon data with the tomographic
statics (if any)
applied. The decomposed solution will be
derived from picks from the entire recorded
volume over the
restricted “good data” time window.
12. RESIDUAL REFLECTION STATICS VELOCITY ANALYSIS
With tomographic (or refraction) statics
and residual reflection statics applied, velocity analyses should
be derived from
the deconvolved data at roughly a half-mile interval in the inline direction
and a quarter-
mile interval in
the crossline direction, utilizing the same interactive QC tools used in the
picking of
preliminary velocities. Filter tests may be performed on the data
input to velocity analysis to determine
the bandpass
filter that produces the highest resolution semblances for velocity picking. That filter, along
with an RMS gain,
should be applied to the input data, but NOT used on the final stacked
volume.
13. REFLECTION RESIDUAL STATICS (2ND PASS)
With tomographic (or refraction) statics, 1st-pass
reflection statics and post-statics velocities applied, a
second pass of
reflection statics should be performed.
For 2nd-pass statics, time deviations will be picked
against a stacked
model in the same “good data” time window, with a reasonable (24ms?) shift
limit. For
pick decomposition, the near Shot-Receiver (e.g., 8000’ for
targets below 8000’ depth) data should be
used. Both the input data and the model should be
RMS gained, but neither should be bandpass filtered.
14. COMMON MIDPOINT FINAL STACK
With all statics and normal moveout applied, the CMP Amplitude
Adjusted volume in Step (9) should
be stacked using
client-approved outside mute parameters.
An example follows:
D3DSP Processing Guide Page
3 of 3 4/11/2002
OFFSET TIME
700’ 70 ms
3950 600
5500 1200
8000 2000
10000 3000
12000 4000
20000 6000
15.
POST-STACK TIME MIGRATION
Migrate with a suitable one-pass 3D time migration program, or apply a cascaded series of migrations
using a minimum function for the one-pass migration and residual laterally varying velocities for two
passes
(Inline and Crossline) of time-domain migration, and again balance the
wavelet's spectrum. Using
Time-Depth or Time-Velocity information
provided by the client, a 3-D migration velocity T’ube should
be created, from which, a single minimum velocity function can be extracted which can be used in the
one-pass
time-domain migration. The full, laterally varying, migration
velocity field and the minimum
velocity function
can then be used to generate the residual velocity field. Using this client-approved
velocity T’ube,
2-D finite-difference residual migration may be run on the one-pass-migrated
data, in the
inline
direction. This volume can then be
sorted into the crossline direction, and a second pass of finite-
difference
migration should be performed.
16.
D3D-REFLECTIVITY VOLUME - (2nd PASS TVSB)
A second pass of time-varying spectral
balancing will be applied post-migration.
The amplitude
spectrum should be
whitened using the same partameters as in Step (3),
above. Zero-phase bandpass
filters are
recommended for spectral decomposition.
This is referred to as the D3D-reflectivity Volume.
17. D3D-IMPEDANCE VOLUME (RUNNING-SUM-INTEGRATION
or INVERSION)
A final impedance volume should be
generated through trace integration of the D3D-migrated TVSB
data. This trace-by-trace operation is a simple
running sum (accumulation) of samples down a tracel,
and is to be
followed by a (~3 Hz?) low-cut filter to remove any DC component that might be
generated
by the first few
(high amplitude?) samples on each trace.
A filter test panel is recommended
to allow the
client to select the low-cut parameters.
This (or the next step) is called the D3D-impedance
Volume.
18. PHASE ROTATION
At the client's request, for polarity
convention purposes, both the reflectivity and the impedance volumes
may be phase
shifted 180-degrees prior to output and delivery.
19.
DELIVERABLES
(Inline & Crossline Numbering Consistant with
Original-processed Data)
SEG-Y - 8mm cartridge of Final, Unfiltered (D3D-Migrated) Reflectivity Volume
SEG-Y - 8mm cartridge of Final, Low-cut Filtered (D3D-Migrated) Impedance Volume
SEG-Y – NMO- and Static-corrected, Unmuted, Prestack CMP Gather Traces
SEG-Y - Final, Unfiltered D3D-Stack (Relative Amplitude)
ASCII - 3D Stacking Velocity Field
ASCII - 3D Time-Depth Field (client supplied)
ASCII - 3D Total Shot and Receiver Statics Field
Paper - Contoured Stacking Velocity Cross-sections and Time Slices, at sparsely selected intervals.
Paper - Final, Fully-corrected CMP Gather Plots (ev 5th CMP on ev 20th Inline, at 2 IPS & 40 TPI)
Paper and Electronic (MS Word) - Processing Report
* Diagnostic 3D Seismic
Process – Patent # 5,671,136, issued 9/27/97,
Licensed to EPL (NOLA) by VTV, Inc. (Denver,
CO)