Docstoc

REMSA PROCESSING TEST NC CGGVeritas Final Report Seismic

Document Sample
REMSA PROCESSING TEST NC CGGVeritas Final Report Seismic Powered By Docstoc
					REMSA– 3D PROCESSING TEST NC186                                    CGGVeritas




               Final Report
          Seismic Data Processing

                3D PROCESSING TEST
                      NC186

                                      LIBYA



                           Period May 2009 – June 2009

                                         For
                                      REMSA
                                          By
                                      CGGVeritas


                     Written by                Checked by            Approved by
     Name :      Stanislaw Warzocha            JL Rivault           Lavdosh Bubeqi
   Position :      Project Leader          Geophysical Advisor   Tripoli Centre Manager
       Date :        June 2009                 June 2009               June 2009
  Signature :




                 Tripoli Processing Center – 3/23/2011                                    1
REMSA– 3D PROCESSING TEST NC186                                                                                                   CGGVeritas



                                                                  CONTENT
1.       INTRODUCTION ........................................................................................................................... 3
     1.1.         SCOPE OF REPORT................................................................................................................... 3
     1.2.         PURPOSE AND OBJECTIVES OF PROCESSING ............................................................................ 3
     1.3.         LAYOUT OF REPORT ............................................................................................................... 4
2.       DATA ACQUISITION ................................................................................................................... 5
     2.1.         SURVEY LOCATION MAP ......................................................................................................... 5
     2.2.         DATA ACQUISITION PARAMETERS .......................................................................................... 6
3.       PROCESSING TESTS ................................................................................................................... 9
     3.1.      MERGE OF 2 SURVEYS & MINIMUM PHASE FILTER DERIVATION ............................................. 9
     3.2.      PRIMARY STATICS .................................................................................................................. 9
        3.2.1.    Primary statics ................................................................................................................ 15
        3.2.2.    First break picking.......................................................................................................... 16
        3.2.3.    Inversion ......................................................................................................................... 16
        3.2.4.    Static solution ................................................................................................................. 17
        3.2.5.    Residual statics ............................................................................................................... 17
        3.2.6.    Conclusion and recommendation: .................................................................................. 18
     3.3.      LINEAR NOISE ATTENUATION (GROUND ROLL) ..................................................................... 19
     3.4.      PRE-PROCESSING SEQUENCE ................................................................................................ 19
        3.4.1.    Frequency Dependant Noise Attenuation ....................................................................... 20
        3.4.2.    Surface consistent amplitude processing ........................................................................ 20
        3.4.3.    Inverse Q filtering ........................................................................................................... 20
        3.4.4.    Surface consistent deconvolution.................................................................................... 21
     3.5.      VELOCITY ANALYSIS ............................................................................................................ 22
     3.6.      SIGNAL TO NOISE ENHANCEMENT BY RADIAL MIXING ......................................................... 23
     3.7.      DMO/DMO-1 ...................................................................................................................... 24
        3.7.1.    Forward Dip Move Out on offset cubes .......................................................................... 24
        3.7.2.    Creation of missing traces on a 3D regular grid ............................................................ 24
        3.7.3.    Attenuation of the footprint acquisition .......................................................................... 25
        3.7.4.    Random Noise Attenuation ............................................................................................. 25
        3.7.5.    Reverse Dip Move Out on offset cubes ........................................................................... 25
     3.8.    PRE-STACK TIME MIGRATION ................................................................................................ 27
     3.9.      RESIDUAL MOVE OUT CORRECTION .................................................................................... 28
     3.10.     DE-MULTIPLE TESTS IN TAU-P DOMAIN ................................................................................ 29
     3.11.     MUTES TESTS ....................................................................................................................... 30
     3.12.     POST PROCESSING TESTS ...................................................................................................... 31
        3.12.1.      Attenuation of the footprint acquisition on PSTM stack ............................................ 31
        3.12.2.      Random Noise Attenuation on PSTM stack ............................................................... 31
        3.12.3.      Time-variant band-pass filtering on PSTM stack ...................................................... 31
        3.12.4.      Post-stack scaling on PSTM stack ............................................................................. 31
4.       PROCESSING SEQUENCE ........................................................................................................ 32
     4.1.      PROCESSING SEQUENCE........................................................................................................ 32
     4.2.      FLOW CHART ........................................................................................................................ 35
        4.2.1.   Final Products ................................................................................................................ 36
5.       PERSONNEL ............................................................................................................................... 37
     5.1.         CGGVERITAS PERSONNEL ................................................................................................ 37
6.       CONCLUSIONS ........................................................................................................................... 38
7.       APPENDIX................................................................................................................................... 39
     7.1.         C.D.P. GRID DEFINITION ................................................................................................ 39
     7.2.         TIME SCHEDULE ................................................................................................................... 39
     7.3.         FINAL PRODUCTS SPREADSHEET ........................................................................................... 39




                                  Tripoli Processing Center – 3/23/2011                                                                                    2
REMSA– 3D PROCESSING TEST NC186                                            CGGVeritas



1. Introduction
1.1. Scope of report

This report describes the processing of the 3D REMSA test survey situated in Murzuq
Basin of Libya, license NC186.


The 3D REMSA test represents approximately 117 km2 data, contains 4.5 km2 of full
fold data, using vibroseis as source.
Two different surveys contributed for the test cube and particular attention was paid
for the merging of two surveys.

Coordinates and elevations were supplied in the headers.
The agreed datum plane was 450m ASL and replacement velocity 2000 m/s.

The data were processed by CGGVERITAS for REMSA at its data processing Center
in TRIPOLI (LIBYA) between May 2009 and June 2009.

REMSA supplied to CGGVeritas a copy of original data on 3590 cartridges which
contained SEGY data with geometry on headers.

CGGVeritas supplied to REMSA the final products on DVD-s, requested QC files
and presentations with different processing steps.


   1.2. Purpose and objectives of processing

The primary objective of the processing is the structural imaging and the fault for the
purpose of bid evaluation

The key points are:
- Improvements in interpretational and structural accuracy by enhancing coherency,
resolution and signal to noise ratio.

- Static model derived from up-holes and/or refraction picking is of high importance.

- De-noising efficiency (linear and random noise). It appears that signal to noise ratio
is very poor in the western and eastern parts of the survey. A big effort should be done
in order to spatially harmonize frequency and amplitude through the survey.




                   Tripoli Processing Center – 3/23/2011                                   3
REMSA– 3D PROCESSING TEST NC186                                              CGGVeritas



   1.3. Layout of report

The report is divided into sections describing the data acquisition (section 2), data
processing tests (section 3), processing sequence (section 4), personnel (section 5),
conclusions (section 6); additional details are given in Appendix (section 7) when
necessary.

Reference to Figures (diagrams within the report text), Plates (A4 size plots grouped
in ppt files) may be made in the processing report.

In appendix 7.3, a table lists the tapes of the final deliveries with their numbers and
content.




                    Tripoli Processing Center – 3/23/2011                                 4
REMSA– 3D PROCESSING TEST NC186                                      CGGVeritas


2. Data acquisition
The processed area concerns the two 3D Murzurq surveys, which cover around 117
km2 of 77 full fold (Acquisition 2003) and 88 full fold (Acquisition 2006).
Corresponding acquisition parameters are described below.


2.1. Survey location map




                 Tripoli Processing Center – 3/23/2011                            5
REMSA– 3D PROCESSING TEST NC186                       CGGVeritas



2.2. Data acquisition parameters
Survey 2003




              Tripoli Processing Center – 3/23/2011                6
REMSA– 3D PROCESSING TEST NC186                       CGGVeritas




Survey 2006




              Tripoli Processing Center – 3/23/2011                7
REMSA– 3D PROCESSING TEST NC186                       CGGVeritas




              Tripoli Processing Center – 3/23/2011                8
REMSA– 3D PROCESSING TEST NC186                                             CGGVeritas


3. Processing tests
This paragraph describes the validation of parameters used for the 3D NC186
processing and different tests performed to establish the optimum processing
sequence; some steps are conventional ones, and some steps are specific to this
processing.



3.1. Merge of 2 Surveys & minimum phase filter derivation

Before starting any processing CGGVeritas made sure that the 2 datasets were ready
to be processed, which the following attributes matched:
    - Phase
    - Amplitude
    - Elevation
    - Signal and noise levels
        After careful QC of above was decided that the 2 datasets have the same
        character and they can easily merge together

The vibroseis data require a conversion to minimum phase before deconvolution: the
minimum-phase operator computation is based on the assumption that the auto-
correlation of the pilot sweep (extracted from auxiliary traces) is a zero-phase wavelet
emitted by the vibroseis source.
The vibroseis source is a long-duration sweep signal in the form of a frequency-
modulated sinusoid that is tapered in both ends.
The correlated vibroseis data is a mixed phase signal due to the nature of the source
(the filtering induced by the earth is minimum phase).
A minimum phase conversion of vibroseis signal is necessary to retrieve the minimum
phase hypothesis, which is the input to the deconvolution.
A minimum-phasing operator with a spectrum amortization is computed from a sweep
autocorrelation, moving from a zero-phase signal to a minimum-phase one.
The polarity of field data was revered based on Remsa’s request

Please refer to the file T001 CGGV_NC186_Merge_MinimumPhasing.ppt for the
results of this step.


3.2. Primary statics

Introduction
The purpose of static corrections is to reduce the observed travel times to what they
would be if the irregularities of the surface of the ground and of the thickness of the
low velocity-weathering layer were not present. That is, to remove time distortion of
the seismic data, resulting from variations in topography, velocity and thickness
variations in the near-surface layers.




                   Tripoli Processing Center – 3/23/2011                                   9
REMSA– 3D PROCESSING TEST NC186                                              CGGVeritas


An important issue in estimating shot and receiver statics is the accuracy of the results
as a function of wavelengths.

The short wavelength static shifts (less than a spread length) cause travel time
distortions in CMP gathers, and thus degrade the stacking quality. However, merely
improving the stack response by correcting for short-wavelength statics may not be
sufficient. That is, the stack response as a criterion for a good static approach is not
enough.

Residual statics corrections are needed because field statics and datum corrections
almost never totally compensate for the effects of near-surface velocity variations.
This is because the near-surface velocity variations are not known and therefore exact
corrections cannot be made. The method of surface-consistent statics estimation based
on reflections works well in accommodating short-wavelength variations, but
performs poorly in handling the long wavelength variations, i.e. surface-consistent
residual statics corrections cannot solve for the long wavelength statics component
(more than two spread lengths).

Refraction static methods are based on first-break arrival times, and, in theory, are
able to estimate the long-period statics component.

The problem is that we need to define a near-surface model, which could comprise
more than one layer.

In an iterative, model-based approach, the computed travel times from an assumed
initial near-surface model are compared with the actual first-break picks. The
procedure minimizes the observed difference in an iterative way.

Datum planes
Description of the different datum planes used for this processing:

1 – Fixed seismic datum plane

The fixed datum to which the seismic data is corrected is chosen locally in order to
avoid static correction with too big values. For this survey client suggested to use the
datum of 450m asl.

2 – Floating datum plane

Especially for velocity analysis, we want to use times measured as close as possible to
the surface in order to take into account the effect of the layers between the datum
plane and the surface. To do so, primary statics are 3D filtered to separate the high
and low frequency static components.

The high frequency component plus further residual statics are applied to the pre-
stack data. The low frequency component LF is used as auxiliary datum (floating
datum plane), which is the zero time reference for velocities. This LF component is
then applied post-stack, Figure 1.



                    Tripoli Processing Center – 3/23/2011                                   10
REMSA– 3D PROCESSING TEST NC186                                             CGGVeritas




   Figure 1: Floating datum

At the floating datum plane time, a pulse is displayed on stacks images to show the
FDP (velocity starting time).

The difference between the floating datum plane and the fixed datum plane is the low
frequency component static correction and is called regional static correction, Figure
2.




   Figure 2: Regional static correction

The regional static correction is obtained by smoothing the individual static
corrections of all traces, which form a CDP. This regional static is applied post-stack.
The time origin is therefore located on the fixed datum plane.

3 – Additional bulk time shift:

An additional time shift (downwards) of 400 ms is applied to preserve shallow data
possibly located above the datum. The stack data start time is then -400 ms, which
means the datum plane is located at 400ms.

The floating datum plane was derived from the upholes derived model statics and the
additional time shift is 400 ms; the floating datum and additional time shift of 400 ms
are used until the end of processing; so all the seismic will be about 400 ms deeper
than the original data.


                   Tripoli Processing Center – 3/23/2011                                   11
REMSA– 3D PROCESSING TEST NC186                                             CGGVeritas




                                         Land Data




                                       SP                Re




                                                                         STATIC
                                                                         PROBLEM?
                           t

   Figure 3: Statics Problem

The following figures 4 to 9 illustrate the concept of floating datum:


                                         Floating Datum Plane



                                                      Elevation
                                                                   Static
                           FDP                                          s
                                                                          WZ         Fixed Datum
                                                                                           Plane
                                                                                           (AWD)




                                                                     FDP=Fn (elevation, wz)

                                                                         Static = SFDP + S Res
           t


   Figure 4: Statics




                       Tripoli Processing Center – 3/23/2011                                     12
REMSA– 3D PROCESSING TEST NC186                                          CGGVeritas



                                   Floating Datum Plane



                                                  Elevation
                   FDP

                                                                    WZ




             t                                                FDP=Fn (elevation, wz)



  Figure 5



                                   Floating Datum Plane


                                             Re          SP
                                                  Elevation
                     FDP




                                                              CDP



             t



  Figure 6




                 Tripoli Processing Center – 3/23/2011                                 13
REMSA– 3D PROCESSING TEST NC186                                              CGGVeritas



                                       Floating Datum Plane


                                                 Re         SP
                                                      Elevation         Re
                       FDP            SP




                                           CDP                    CDP



          t



  Figure 7: Floating datum.



                                       Floating Datum Plane


                                   With Datum Plans below elevation
                                                Re        SP
                                   SHALLOW DATA above zero Time
                                                   Elevation        Re
                       FDP             SP         is lost

       t=0                                                                      Fixed Datum Plane




                                           CDP                    CDP



          t



  Figure 8: Shallow data is lost




                   Tripoli Processing Center – 3/23/2011                                      14
REMSA– 3D PROCESSING TEST NC186                                              CGGVeritas




                                     Floating Datum Plane

                               Shift Down
                               to preserve data above datum


      t=-x                                       Re         SP
                                    SP                Elevation         Re
                       FD
                       P
      t=0                                                                           Fixed Datum
                                                                                    Plane




            t                              CDP                    CDP

  Figure 9: Shift down to preserve data above Datum.


All elevations in Libya are referenced to the Mean Sea Level (ASL).

3.2.1. Primary statics

The following provides a description of the trials of primary and residual static
computation performed in this processing project.

Several solutions to estimate primary statics were tested and compared.

       - Elevation statics
       - 3D weathering model (based on upholes acquired with the 3D) including dune
       modeling (2 different models were tested, 1 layer and 2 layers)
       - 1 layer refraction statics (SDITR for primary statics)
       - Full refraction statics (Geostar (equivalent to GLI Generalized Linear Inversion))
       once with automatic FB picks and once with SDITR FB picks.
       - Full refraction statics using tomographic inversion
       - Residual Refraction Statics after uphole model statics


Primary statics methodology using up-holes information

Based on the upholes, a near surface model was built.
The model is constituted of 2 layers.
The first layer is composed of dune model and a thin layer below surface (upholes).
The second is coming from upholes and it goes down to the base of dunes.




                   Tripoli Processing Center – 3/23/2011                                      15
REMSA– 3D PROCESSING TEST NC186                                             CGGVeritas


Primary refraction statics methodology

All primary statics solution except up-hole model, are based on refraction. In parallel,
elevation statics, to only correct for elevation, are generated to produce a reference
stack. Associated QC stacks are produced and in addition residual statics are tested to
correct for the remaining anomalies.

Three methods have been tested:
-      SDITR application:
SDITR is using a tomographic approach to invert the forward and reverse travel times
from stacks (first break stacks) and generates a near surface layered model.
-      Geostar application (Generalized Linear Inversion):
GLI has been described by Hampson & Russell in 1984. It provides a near surface
depth/velocity model after inversion of the picks performed on elementary traces.
Two methods have been tested:
                - Blocky model (2 layers) (linear inversion)
                - Tomographic model (tomographic inversion), not successful result.


3.2.2.First break picking

First break picking and our statics solutions were focused on the refractor coming on
the offsets between 1200 to 1800 m.

A minimum phase conversion was first performed, and then the picking was carried
out

 Using first break stacks provides a quick and robust picking, very effective when
noise is polluting the first breaks (see common receiver refraction stack on which the
refractor at around 200 ms is picked: red and green stacks correspond to forward and
reverse travel times).
This first picking was then used to guide the picking of the elementary traces.
The first break picks were of poor quality is some areas, and then the solution based
on first breaks was not reliable.


3.2.3. Inversion

-        Geostar:
Geostar updates a near surface depth/velocity model by inversion of the refraction
travel times. First, an initial model is defined: V0 was set to 1000 m/s. Then an
iterative inversion process progressively updates the model by minimizing the
difference between the actual picked refraction travel times and the model-based
travel times derived from a refracted ray-tracing realized on current model.

The linear inversion leads to a layered model ("blocky" model) made of
individualized layers with a horizontally variant but vertically constant velocity. For
this project a 2 layer model was tested.




                   Tripoli Processing Center – 3/23/2011                                   16
REMSA– 3D PROCESSING TEST NC186                                             CGGVeritas


The tomographic inversion leads to horizontally and vertically variable velocity
layers.

-    SDITR:
SDITR uses a similar process but based on travel times from stacks.


3.2.4. Static solution

All the different solutions were compared. It was decided that the up-hole derived
model followed by residual refraction statics gave the most reliable solution.
Shot and receiver statics were computed on the updated model.
The static solution was produced using a replacement velocity of 2000 m/s and the
ASL as the reference datum plane.
The medium to long wavelength was reasonably well solved.
With SDITR, refraction statics can be refined by reiterating the first break picking
using the first break stacks performed after application of the first pass of primary
statics.
First breaks were re-picked and a set of residual refraction statics was generated.
Improvements coming from the residual refraction statics are significant; the
continuity of horizons was enhanced.
We believe that more work can be done with the first break picking, including but not
limited to a better conditioning of first breaks before picking, combination of manual
and automatic picking, if we have to get better results from refraction statics (GLI)
The use of SDITR residual refraction statics yields better results because the first
breaks are picked on shot and receiver stacks

Please refer to the file T005_CGGV_NC186_Primary_Statics.ppt for the results of
this step.

3.2.5. Residual statics

 3.2.5.1. Medium and Short wavelength automatic residual static: TDSAT


Surface consistent residual statics using reflections have been computed. These statics
have been computed after all pre-processing steps including noise attenuation, surface
consistent amplitude corrections, phase and amplitude deabsorption and
deconvolution.
The 3D automatic surface consistent medium wavelength residual statics were
computed in one pass for the whole survey.
The criterion is the optimization of CDP stacks, and the program determines the set of
surface consistent time shifts (static corrections), which generate the optimum stack.
The program will not correct large static anomalies or long wavelength statics.
An automatic residual static program TDSAT was tested using a medium vector of
400 meters (32 bins) in the 2 directions. Used with a large gate (300-2000ms), we can
expect to correct for anomalies that affect on the vector length, all the horizons in the
computation gate.
The results were really significant. Continuity was clearly better and the signal to
noise ratio improved.




                   Tripoli Processing Center – 3/23/2011                                    17
REMSA– 3D PROCESSING TEST NC186                                              CGGVeritas


After revision of the velocities, a second pass of surface consistent residual statics was
computed with a shorter vector of 250 meters (20 bins) in the 2 directions.
There were again improvements in some places. It was decided to keep the 2 passes of
surface consistent residual statics.

3.2.6. Conclusion and recommendation:

Using the up-hole derived model for primary statics followed by residual refraction
statics computed with SDITR, plus surface consistent residual statics, we have been
able to provide a static solution that removes anomalies and improves the stack.
It has been noticed that the current solution is optimum in the dune area. It was
assumed that dunes were only composed of sand, however, it seems they are possibly
composed of sand covering hard rock. The decision was taken (with deadlines in
mind) to keep the current model, as the dune area. For future re-processing, a more
complicated static solution model over the dune area could be envisaged.

Please refer to the file T006_CGGV_NC186_Vels&ResidualStatics.ppt for the
results of this step. (Combination of velocity analysis and the residual statics)




                   Tripoli Processing Center – 3/23/2011                                     18
REMSA– 3D PROCESSING TEST NC186                                            CGGVeritas



3.3. Linear noise attenuation (ground roll)

The FKxKy filtering (cone velocity rejection) was tested to remove the source noise
in the cross spread domain.

The 3D FK is working on a cube of data regularly sampled, provided either by one
receiver line and one shot line or by a combination of receiver lines and shots.
On this survey, a cross-spread is constituted with VPs using the same active part of a
receiver line.

The linear noise velocities (ground roll) were measured at 300-1800 m/s.
First-break repetition 2400 m/s (2000-2500 m/s)

Several velocity fans were tried:
 2000,3000 m/s, 24dB, 36dB, 48dB attenuation.
 2500,3000 m/s, 24dB, 36dB, 48dB attenuation.
 2500,3500 m/s, 24dB attenuation
 2250,3000 m/s, 36dB attenuation

Parallel to the 3D filtering on cross spread CGGVeritas tried the latest program called,
AGORA-Adaptive Ground Roll Noise Attenuation, which is used to attenuate the
linear noise and the guided wave

Please refer to the file T002_CGGV_NC186_AGORA_Tech.ppt for a technical
description of AGORA program

QC on gathers of near, midle, far traces and also on stacks was performed.
The frequency spectrums were checked as well.

Based on those tests, we decided to apply the Frequency Dependent Noise
Attenuation (FDNAT) and 2 passes of AGORA, the first one for the linear noise
attenuation and after that on cascade we applied AGORA for attenuation of
guided wave

Please refer to the file T003_CGGV_NC186_AGORA_XS_Results.ppt for the QC of
Agora results on cross spreads

 Please refer to the file T004_CGGV_NC186_SP_Results.ppt for the QC of Agora
results on shot point

Please refer to the file T009_CGGV_NC186_STACK_Results.ppt for the QC of
Agora results on stacks


3.4. Pre-Processing sequence

This processing step is to achieve good amplitude compensation prior to the surface
consistent deconvolution; it includes a surface consistent amplitude correction (source
and receiver).


                   Tripoli Processing Center – 3/23/2011                                   19
REMSA– 3D PROCESSING TEST NC186                                                CGGVeritas


3.4.1.Frequency Dependant Noise Attenuation

The data was highly contaminated by air wave noise, which has very high amplitude.
A new algorithm (FDNAT) was applied to attenuate such noise.
Using frequency-dependant and time-variant threshold values of sample amplitudes,
with defined trace neighborhood, FDNAT detects and suppresses specific noise to
different frequency ranges and different time.
The airwave was efficiently removed.

3.4.2.Surface consistent amplitude processing

This processing step is to achieve amplitude compensation prior to the surface
consistent deconvolution.
The technique computes, within a given window and in one pass, 3D amplitude
corrections for each trace, expressed in centibels with respect to a reference level.
Calculations include source and receiver terms and spatial filtering capabilities. A
Gauss-Seidel decomposition is used. A by-product of this process is the detection of
noise-polluted traces.

                                                                                   A 
For every trace (T), a gain (G in centibel) was calculated. G T   200  log 10       
                                                                                   5000 
where (A) is the average amplitude in a time window of 400-2400 ms.

The surface consistent amplitude compensation is the only way to compensate for
amplitude variations caused by the near surface conditions.

Surface consistent amplitude compensation is run on minimum phase dataset.

This compensation is applied after FDNAT and 2 passes of AGORA production and
spherical divergence correction (TV2) through three steps:

   Amplitude compensation versus offset
   Edition of traces with anomalous amplitude
   Surface-consistent amplitude correction (Source & Receiver)

3.4.3. Inverse Q filtering

The Q factor was estimated after FDNAT and 2 passes of AGORA on a stack In-line
without deconvolution, without filter by measuring the spectral absorption between
the signal spectra extracted from two time windows:
    - A shallow window: T= 500-1500 ms.
    - A deeper window: T=1500-2500 ms.

As QC of the measurement Q136, two stacks were computed without deconvolution
applied or filtering:
   - A reference stack without Q compensation.
   - A stack with Q compensation in phase and amplitude.




                    Tripoli Processing Center – 3/23/2011                                    20
REMSA– 3D PROCESSING TEST NC186                                           CGGVeritas


The regional value used by CGGVeritas (Q136) was confirmed with this
measurement.

The de absorption in amplitude increases the high frequency content and provided a
sharper signal. In a few places, the noise was slightly boosted, but at a reasonable
level. The noise can be removed later.

3.4.4. Surface consistent deconvolution

The tests were performed after Phase and Amplitude Q136 compensation.
Different surface consistent deconvolutions were tested with the following parameters

   -   Spiking, single gate 200-2000 ms, 120 ms operator length, 3% white noise.
   -   Spiking, single gate 200-2000 ms, 160 ms operator length, 3% white noise.
   -   Spiking, single gate 200-2000 ms, 200 ms operator length, 3% white noise.

   -   Gapped, single gate 200-2000 ms, 160 ms operator length, gap 16 ms, 3%
       white noise.
   -   Gapped, single gate 200-2000 ms, 160 ms operator length, gap 24 ms, 3%
       white noise.
   -   Gapped, single gate 200-2000 ms, 160 ms operator length, gap 32 ms, 3%
       white noise.

The spiking deconvolution was significantly increasing the noise level.

The predictive deconvolution reduced the ringing that occurred from place to place.

The 160 ms operator length with a gap of 32ms gave the sharper image.


Please refer to the file T004_CGGV_NC186_SP_Results.ppt for the QC of the above
step results on shot point

Please refer to the file T009_CGGV_NC186_STACK_Results.ppt for the QC of these
steps on stacks




                   Tripoli Processing Center – 3/23/2011                                21
REMSA– 3D PROCESSING TEST NC186                                            CGGVeritas


3.5. Velocity analysis

A first pass of 1500mX1500m velocity analysis was done. The velocity field was
accurately checked by the team leader. The velocity field was well correlating with
expected structures.

After this step, an automatic surface consistent residual statics computation was
performed using TDSAT program

A second pass of 750mX750m velocity analysis was done. The velocity field was
accurately checked by the team leader. The velocity field was well correlating with
expected structures.

After this step, an automatic surface consistent residual statics computation was
performed using TDSAT program

Please refer to the file T006_CGGV_NC186_Vels&ResidualStatics.ppt for the
results of this step. (Combination of velocity analysis and the residual statics)




                   Tripoli Processing Center – 3/23/2011                                22
REMSA– 3D PROCESSING TEST NC186                                           CGGVeritas




3.6. Signal to Noise Enhancement by radial mixing

The signal to noise ratio was very poor in some
Standard methodology such FX filtering was not efficient as signal was not strong
enough to be distinguished from noise in such areas.
CGGVeritas proposed a new algorithm RADMX, which is a radial stack within a
Cross spread. Done on NMO corrected Cross spread gathers, radial stacks are
generated for each trace of the gather. The radius was of 25 meters. The current trace
is then replaced by the local stacked trace. Doing it in Cross spread domain
guarantees a good preservation of azimuth and offset information.
The results are relatively spectacular by improving significantly the signal to noise
ratio without loosing fault definition and high frequency content.
The benefit of RADMX was confirmed by a test of PSTM done with and without
RADMX.

Please refer to the file T007_CGGV_NC186_RadialFiltering.ppt for the results of
this step.

Please refer to the file T003_CGGV_NC186_AGORA_XS_Results.ppt for the
results of this step on cross spreads (time slices)




                   Tripoli Processing Center – 3/23/2011                                 23
REMSA– 3D PROCESSING TEST NC186                                              CGGVeritas


3.7. DMO/DMO-1

Cross-spread acquisitions have a strong pattern imprint due to clear offset distribution
pattern and poor contribution of near to medium offsets in comparison of medium, far
offsets.
In the last slide of file T007_CGGV_NC186_RadialFiltering.ppt we explained 3
different methodologies for bin centering and offset regularization of this dataset.

CGGVeritas would normally propose the latest processing techniques as COV
(Common offset vector) or full 5D interpolation techniques in order to bin-center and
regularize the data.

Due to the time constraint we ran an offset regularization including bin centering
through the DMO/Reverse DMO sequence.

An AMO (Azimuth Move Out) process, in offset cubes, before the Pre-Stack Time
Migration was tested for fold regularization and noise attenuation:

          Dip Move Out (azimuth effect removed, data is bin centered)
          Regularization (missing trace interpolation)
          Acquisition footprint attenuation
          Random Noise Attenuation
          Reverse Dip Move Out


3.7.1. Forward Dip Move Out on offset cubes

Several scenarios for DMO offset slotting were proposed.
The first one was to create 29 offset classes without holes, preserving enough offset
classes for RNMO corrections.
The Dip Move Out was performed using offset slots of 75 meters (except for the first
classe (0-150m), and the two last ones giving a total of 29 offset .

The acquisition geometries induce in places large irregularities in the offset
distribution per bin: trace weighting per offset slot is applied to minimize the effect of
these irregularities. Traces of low weight were dropped after DMO.

3.7.2.Creation of missing traces on a 3D regular grid

The module MISTR creates the missing traces in a 3D block.
The calculations are performed in the f-x-y domain as follows:
    Determination of a prediction operator from the existing traces.
    Calculation of the missing traces from the existing traces and the prediction
       operator. Only a maximum of three missing traces have been interpolated both
       side of gaps. This process was done for all the offset classes




                    Tripoli Processing Center – 3/23/2011                                    24
REMSA– 3D PROCESSING TEST NC186                                              CGGVeritas


3.7.3. Attenuation of the footprint acquisition

The module FKF3D used to attenuate acquisition imprint, can detect and subtract
acquisition footprint from the input volumes. The input data is converted from the
time-space domain to the frequency-space domain to obtain an (f-x-y) cube where f is
the temporal frequency, X is the Cross-Line and Y the In-Line number. Acquisition
footprint is an additive noise with spatial periodicity. This periodicity shows as
spectral peaks when transformed to the wave number domain (f-kx-ky). Attenuation
energy at these peaks and inverse transforming to the f-x-y (and then to t-x-y) provide
footprints attenuated data. This process was done for all the offset

3.7.4. Random Noise Attenuation

The module PRF3D used to attenuate the Random Noise, by projection filtering in
3D, performs a projection filtering in the (f-x, ky) domain.
This module separates the signal, assumed to be predictable in x, from non-
predictable noise, for all the signal’s component frequencies. A projective filtering is
used; the signal after filtering is the same as it would be after passing through a filter
whose spectrum values are 0 or 1. This ensures that the signal is preserved at the same
time as optimizing the attenuation of random noise. The projective filter is calculated
from an auto-deconvolved predictive error filter. This process was done in the same
offset classes described above.

3.7.5. Reverse Dip Move Out on offset cubes

The 3D DMO is used for the offset cube regularization and summation. Summing the
multi azimuth data to produce single fold offset cubes requires taking into account the
azimuthal variation within the data. A band limited spatial interpolation is used to
perform the bin centering. The bins along the source receiver axis are illuminated.
For each bin and each offset, the inverse DMO restores a shot and receiver position
with the main acquisition azimuth.

Please refer to the file T008_CGGV_NC186_OffsetRegularization.ppt for the results
of the above steps.




                    Tripoli Processing Center – 3/23/2011                                    25
REMSA– 3D PROCESSING TEST NC186                       CGGVeritas




                                   Figure 3.9




              Tripoli Processing Center – 3/23/2011                26
REMSA– 3D PROCESSING TEST NC186                                            CGGVeritas




3.8. Pre-stack Time Migration

CGGVeritas solution for pre-stack time migration is a Kirchhoff pre-stack time
migration type (TIKIM). The algorithm is a trace-by-trace migration which treats
each output sample as the apex of a diffraction curve defined by the migration
velocities and the true source and receiver positions (including elevations). The
kinematics of the rays contributing to a reflection event are based on a 1D
assumption, i.e. the ray trajectories to a migrated position will only depend on the
velocities at this position and not on the surrounding locations. Amplitude artifacts
introduced by irregularities in the acquisition geometry can be partially compensated
for by the use of a weighting scheme based on a stationary phase zone analysis.

The Pre-Stack Time Migration is performed using CDP gathers after reverse DMO.
The idea is to produce migrated images (for stack and CDP gathers) using
perturbations of a reference velocity function.
To derive the migration velocity function, a first pass of PSTM is run to output the
velocity lines as well as gathers, fully migrated with a range of trial velocities: 16
fully migrated stacks were produced for each velocity lines using 16 different velocity
fields, which correspond to the perturbations of the previous velocity field (NMO
velocity field), 95% to 110%, each 1%. The velocity grid was of 500m x 500m.
Pre-Stack Time Migration velocity picking was performed with the Chronovista tool.
The main advantages of Chronovista for PSTM velocity picking are:
             To allow picking perturbed velocity directly on migrated stack images.
             To validate the picking, by flattening these events, on the CDP gathers.

From this, a final migration velocity field was obtained, after appropriate smoothing.
All offset cubes were migrated independently.

Dip limit tests and aperture tests were done. With a fixed 5km aperture, the following
dip limits were tested 20°, 25°, 30°, 35°, 40°, 50° and 60°. At the target level of
around 1000/1100ms the dip limit of 30° was better than 25°. The dip limit was
changeable with increasing time:
       T        0 ms           dip limit 10°
       T        800 ms         dip limit 20°
       T        1500 ms        dip limit 30°
       T        3000 ms        dip limit 30°
       T        4000 ms        dip limit 20°

For comparison the Fast Track used:
Time 0ms - Dip 10°, T600ms – Dip 20°, T1600 – Dip 35°, T3000 - Dip 35°, T4000 –
Dip 20°.
The dip was kept constant at 30° and the following migration apertures were tested; 2,
2.5, 3, 3.5 and 4km. It was decided to use an aperture of 2000 meters and a time
variant dip limit as following:




                   Tripoli Processing Center – 3/23/2011                                  27
REMSA– 3D PROCESSING TEST NC186                                             CGGVeritas


3.9. Residual Move Out Correction
After pre-stack time migration, a residual move out correction was estimated, using
the CGGVeritas algorithm HDPIC4. A high density automatic picking done on
gathers generates dense volumes of velocity and anellipticity moveout parameters. It
is based on an original approach of an-elliptic shifted hyperbola moveout equation
which corresponds to a 6th order approximation of anisotropic travel times. Both
velocity and anellipticity parameters are simultaneously estimated using the full
coverage of the data. The scans of two internal uncorrelated parameters generate bi-
spectral panels at any time level. The sampling of these parameters is directly related
to the sensitivity of the normal moveout.
For practical reasons, HDPIC outputs the uncorrelated parameters dtn and τ0 to help
further interpolation and filtering of moveout parameters.
A 3D corridor limited by the minimum and maximum values of velocity and
anellipticity attempts to reduce the analysis domain. The maximum of the semblance
is used as the automatic picking criteria.
The outputs of HDPIC are related to the following parameters:
1) DN: (dtn, which is related to the velocity, η and t0)
2) TO: (τ0, which is related to η and t0)
3) Semblance (0-100)

The two attributes dtn and τ0 are then filtered independently. A filter of 250 meters
was finally used. Final filtered dtn and τ0 are converted into velocity and an-ellipticty
field.
The flattening of the gathers was clearly improved.




                   Tripoli Processing Center – 3/23/2011                                    28
REMSA– 3D PROCESSING TEST NC186                                              CGGVeritas



3.10.De-multiple tests in Tau-p domain
The aim of this HR Radon processing was more to clean the gathers rather than to
attenuate multiples on the stack.
The technique requires setting a limit between primary reflections and multiples on a
NMO corrected gather.
This limit is implementing in giving a residual NMO value (rNMO corresponding to
parameter DTCUT) at the maximum offset XRM: 2500 m.
See below the definition of the parameter DTCUT.


                                                            3200 m


                                                                                               -240 ms




                                                                                           Primary
                                                                                          after NMO
                                                                                      Tested




                                                                                           Multiple
                                                                     16 ms                after NMO




  The target is assumed to be at 1250 ms and the rms velocity around 2950 m/s.
                                                                                               900 ms

  The following values of the R NMO were tested:

  R NMO = 64 ms at 3800 m

  Several tests of DTCUT were done.

   R NMO = 100 ms at 2500 m
   R NMO = 120 ms at 2500 m
   R NMO = 140 ms at 2500 m
   R NMO = 160 ms at 2500 m
   R NMO = 180 ms at 2500 m
   R NMO = 200 ms at 2500 m
   R NMO = 240 ms at 2500 m
Comment:
The radon de-multiple removes clearly aliased energy on the CDP gathers, but the
improvement on the stack remains limited due to the ability of the stacking process to
reduce noise and multiples.

Please refer to the file T010_CGGV_NC186_PSTM_Demultiple.ppt for the results of
the above steps.


                   Tripoli Processing Center – 3/23/2011                                         29
REMSA– 3D PROCESSING TEST NC186                                              CGGVeritas


3.11.Mutes tests

Some mute tests were done to obtain the optimum image.
Angle stacks were generated.
The final mute for the full stack was fixed at 40° based on the final velocity field.




                    Tripoli Processing Center – 3/23/2011                                 30
REMSA– 3D PROCESSING TEST NC186                                         CGGVeritas



3.12. Post processing tests


3.12.1. Attenuation of the footprint acquisition on PSTM stack

On PSTM time slices some more imprint acquisition (light) could still be observed, so
we applied on stack the same process that the one which was applied for each offset
classes and described in the paragraph 3.9.3. As expected, the effect is minor, as
remaining imprint was mixed during the reverse DMO process and the PSTM.


3.12.2.Random Noise Attenuation on PSTM stack

The module (COH3D) used to remove random noise attenuation on post stack using
slant stack techniques


3.12.3.Time-variant band-pass filtering on PSTM stack
A time-variant band-pass filtering was applied post-stack.
       Time      0-1200 ms : 8,14,65,90 Hz
       Time 2000-3000 ms : 6,12,50,65 Hz
       Time 3800-4000 ms : 6,10,40,50 Hz


3.12.4.Post-stack scaling on PSTM stack
A post-stack equalization of 1200ms was applied, with 50% overlap between
windows.

Please refer to the file T011_CGGV_NC186_PostStack.ppt for the results of the
above steps.




                  Tripoli Processing Center – 3/23/2011                                 31
REMSA– 3D PROCESSING TEST NC186                                         CGGVeritas



4. Processing sequence

4.1. Processing sequence

Approximately 10 km2 (total surface) of seismic data were processed from SEG-Y on
DVD with the following parameters:

General
           Processing length        : 4000 ms
          Sampling rate           : 2 ms
          Datum plane             : (ASL): 450 m above mean sea level
          Replacement velocity    : 2000 m/s
           Polarity                 : SEG Reverse

           GRID origin was supplied by Remsa


  Reformatting from SEG-Y to CGG format
  Geometry up-dating
  Positioning and reverse traces controlled
  3D Binning 12.5 m x 12.5 m
  Automatic de-spiking
  Elevation statics (DP=450m/s, Vr=1000m/s) removed after 3D FK filtering
  Frequency dependent noise attenuation –FDNAT
  3D linear noise attenuation using AGORA
  Guided Wave Noise Attenuation using AGORA
  Primary statics derived from up-holes model
  (application to floating datum (FDP))
  Residual refraction statics
  (application to floating datum (FDP))
  Spherical divergence correction
   Each sample is multiplied by: T*V2 where: T is the two-way-time
   V is the estimated stack velocity      (Time ms, Velocity m/s):
   (200,2400), (465,2850), (700,3050), (790,3200), (1010,3500), (1410,4000), (2160,4300)
   (3250,4500),
  Regional AVO compensation
  Surface consistent amplitude correction: Source, Receiver, Offset
  Statistical data editing
  Deabsorption in phase and amplitude: Q=136
  Minimum phasing filter on vibroseis
  Surface consistent multi channel predictive deconvolution.
   Operator length: 160 ms, gap: 32ms, pre-whitening: 3%; Gate: 200-2000 ms
  First pass velocity picking
   (1.5 km x 1.5 km grid)
  3D automatic medium wave length residual statics
   Computed in one pass (window 300-2000 ms), vector length: 400 meters
  Second pass velocity picking


                  Tripoli Processing Center – 3/23/2011                                    32
REMSA– 3D PROCESSING TEST NC186                                      CGGVeritas


  (0.75 km x 0.75 km grid)
 3D automatic short wave length residual statics
  Computed in one pass (window 300-2000 ms), vector length: 250 meters
 Radial Mixing in X-spread domain
 Fold harmonization (in offset cubes)
  Forward DMO
 Missing traces interpolation (in offset cubes)
 Acquisition footprint attenuation (in offset cubes)
 Random Noise Attenuation (in offset cubes)
 Fold harmonization (in offset cubes)
 Reverse DMO
 PSTM for velocity lines
  (16 PSTM images with stacking velocity from 95% to 110% each 1%)
  Grid 500 m x 500 m
 Structural velocity picking
  Grid 500 x 500 m
 Filtering of PSTM velocity field
 Full 3D Kirchhoff PSTM
 High density Residual Move Out
 (50 x 50 m grid) filtered at 250m
 High resolution radon demultiple
  R NMO 240 ms WT (200 – 4000 ms), taper 500ms
 Final mute
 40° angle mute based on final velocity function
 Raw PSTM full stack
 Acquisition footprint attenuation on PSTM stack
 Random noise attenuation on PSTM stack
 Final PSTM Volume-Unfiltered and Unscaled
 Spectral whitening on PSTM stack
 Time Variant Filter
  8,14,65,90 Hz                        from      0 to 1200 ms
  6,12,50,65 Hz                        from 2000 to 3000 ms
  6,10,40,50 Hz                        from 3800 to 4000 ms
 Post-stack Scaling
 1200 ms gates, 50% overlap
 Static application form FDP to DP (application of the low frequency component statics)
 Application auxiliary datum DP : 200ms shift (international sign)
 SEG-Y output (Final post-processed Pre-stack time migration full stack)




                 Tripoli Processing Center – 3/23/2011                                33
REMSA– 3D PROCESSING TEST NC186                       CGGVeritas




              Tripoli Processing Center – 3/23/2011                34
REMSA– 3D PROCESSING TEST NC186                                      CGGVeritas



4.2. Flow chart

                        NC186 3D: PROCESSING FLOW CHART


                    Field tape
                  SEG-Y format


                Geometry updating                                      Primary statics
                   3D binning                                           computation
                                                                    Up-holes based model
                                                                              +
                                3D Linear and Guided Noise           Residual refraction
                                       Attenuation                         statics
                                Frequency Dependant Noise
                                       Attenuation

                                     Surface consistent
                                   amplitude computation


                                  Pre-processing (primary
                               statics, amplitude processing)


                                     Surface consistent
                                  Predictive deconvolution

                                 First pass velocity picking

                3D medium wavelength residual statics (one automatic pass)

                                Second pass velocity picking


                         3D automatic short wavelength residual statics


                                           Stack


                                Radial mixing in XP domain


                                        Forward DMO
        Missing traces interpolation, footprint attenuation, Random noise Attenuation
                                        Reverse DMO

               Tripoli Processing Center – 3/23/2011                                    35
REMSA– 3D PROCESSING TEST NC186                                        CGGVeritas




                       PSTM on velocity lines, structural velocity picking


                                   Full 3D Kirchhoff PSTM


                                     High Density RNMO


                              High Resolution Radon Demultiple


Angle Stack                    Final mute & Final PSTM full stack


                                      Footprint attenuation



                                  Random Noise Attenuation


                                       Time variant filter



                                      Time variant scaling


                                                .
                                 Static application from FDP to DP
                   Application auxiliary datum DP (400 ms shift international sign)

                                         Final Products.
                                                  .

4.2.1.Final Products

Based on Remsa’s request CGGVeritas prepared and send to Remsa a DVD
containing all the requested products.
The full content of DVD is explained on the list of final deliverables.




                 Tripoli Processing Center – 3/23/2011                                36
REMSA– 3D PROCESSING TEST NC186                                      CGGVeritas



5. Personnel

5.1. CGGVERITAS personnel

NC186 3D project was produced by:
Missoum Sid Ahmed, Team Leader
Stanislaw Warzocha, Project Leader
Muftah Saad, processing geophysicist

The project was supervised by Lavdosh Bubeqi, Center Manager of
CGGVeritasAgesco Processing Centre and Jean-Louis Rivault, technical supervisor
for CGGVeritas Tripoli.




                  Tripoli Processing Center – 3/23/2011                           37
REMSA– 3D PROCESSING TEST NC186                                            CGGVeritas



6. Conclusions
At each step of processing, information provided by Remsa was very useful.

Several steps of the processing lead to a better data continuity.

Radial mixing in X-spread domain, fold harmonization and regularization of these
multi-azimuth data by using the “DMO/DMO reverse” sequence (forward DMO,
missing trace interpolation, footprint attenuation, random noise attenuation, reverse
DMO) induced an increase of the signal to noise ratio and a reduction of the strong
acquisition footprint, leading to a better data continuity.

The initial result from the fast track was generally good.

The structural image and the fault definitions were improved after the full Pre Stack
Time Migration.

CGGVeritas would have preferred to have more time to test the following:
  1- New techniques for first break picking
  2- COV processing flow
  3- Full 5D Interpolation techniques for full data regularization




                    Tripoli Processing Center – 3/23/2011                               38
REMSA– 3D PROCESSING TEST NC186                                          CGGVeritas



7. Appendix
7.1. C.D.P. GRID DEFINITION

Grid was supplied by Remsa

7.2. Time schedule

The last data was delivered to CGGVeritas on mid June 2009.
The deadline for the delivery of the test results was June 18th, 2009.
The final PSTM volume was delivered on June 18th 2009.


7.3. Final products spreadsheet

Please refer to this file CGGV_NC186_DVD_Content.txt for detailed list of products
delivered on DVD




                   Tripoli Processing Center – 3/23/2011                              39

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:215
posted:3/23/2011
language:English
pages:39