first break volume
23, June 2005
special topic
Multi-disciplinary Geoscience
New concepts for borehole and VSP data integration in seismic data processing Rick Walia and Ken Brandt, CGG Canada Services; Rob Duthie, VSFusion/Baker Atlas; Roger Whittaker and Emmanuel Malterre, First Calgary Petroleums; and Peter Bell, RPS Hydrosearch, provide an example from Algeria to illustrate the benefits of integrating borehole and VSP data in the processing of challenging seismic data.
T
he Berkine basin, Algeria presents unique challenges for seismic data acquisition, processing, and interpretation. This is primarily due to surface and shallow subsurface conditions, namely, namely, the presence of up to 2 km wide and up to 250 m high star-shaped dunes, unconsolidated sands in the shallow subsurface, and several high-density/velocity layers within 500 m below the surface causing severe statics and multiple problems. The seismic data processing was therefore faced with several challenges, in terms of statics, noise (random, linear, multiples and acquisition footprint), and also velocity analysis. These all had a detrimental effect on the integrity of the structural and stratigraphic interpretation of the 2D and 3D seismic data. New ways of integrating borehole data to obtain a better understanding of these problems and possible solutions were considered, and are presented in this article.
(Figure 1). Several early to upper Jurassic salts, anhydrite, anhydrite, carbonates, and evaporates together with unconsolidated sands in the shallow to very shallow subsurface are the source of most of the processing woes in terms of statics, multiples and poor signal-to-noise ratio. For a detailed mapping of the Ledjmet-405/406 blocks, two 3D seismic surveys were acquired using Vibroseis 100fold parallel-geometry. This acquisition geometry allowed a much better sampled offset distribution (50-5000 m) in the CDP domain. This also resulted in effective noise attenuation, improved velocity analysis, an enhanced signal-to-noise ratio, and a stable full Kirchhoff Prestack Time Migration (KPSTM). A typical raw CDP gather before and after noise attenuation and surface-consistent amplitude correction is shown in Figure 2.
Role of borehole data in statics solution Data processing The Berkine basin in eastern Algeria is regarded as one of the more significant hydrocarbon plays in North Africa. The discussion in this paper focuses primarily on the important Ledjmet blocks 405 and 406 of the Berkine basin, which have the potential of more than 5 trillion ft3 of gas and 5 billion barrels of oil equivalent (boe) of recoverable oil reserves
Figure 1 Berkine Basin oil and gas trend - more than 5 billion barrels of oil discovered to date.
© 2005 EAGE
It is obvious from Figure 2 that, due to inaccuracies in first break picking, a conventional statics solution based solely on refraction is bound to fail. This was the first challenge to the
Figure 2 (a) A typical CDP gather from the Ledjmet 3D volume. Poor signal-to-noise ratio due to random noise, ground roll, surface-generated noise, interbed multiples, and refractions produced by several high-velocity sand, limestone, and anhydrite layers posed the main challenges in the surface-consistent amplitude-preserving processing flow. (b) CDP gather after application of random noise and linear noise attenuation in shot and offset domain and multichannel median filters in selected frequency bands.
83
special topic
first break volume
23, June 2005
Multi-disciplinary Geoscience
Figure 3 Combination of model-based layer stripping and reflection statics led to a stable statics solution. Yellow dots in (A) represent upholes and red dots represent exploration wells.
by shallow sand layers. The application of this long-wavelength statics permitted accurate picking of the first breaks corresponding to the deeper ~2000 m/s and ~3000 m/s layers at a modelled depth of ~100 m and ~320 m below ground level respectively. Refraction statics were then computed based on this model as step two of the statics workflow and applied. This step was able to handle medium to long wavelength statics. As step three, a reflection-based specialized algorithm, which uses a known shallow reflection as a reference to compute residual medium wavelength statics, was applied successfully followed by three passes of residual statics (Figure 4). After every pass, velocity picks were refined and the frequency bandwidth was broadened to achieve a high-resolution seismic volume. The final stacks are compared in Figure 5.
High resolution VSP borehole data integration process. Several weeks were spent on experimenting to find the best possible way to handle statics due to surface dune topography, unconsolidated/wet shallow sand layers, and deeper high-velocity laterally-varying thin layers. As a result, a four-step statics workflow was designed as shown in Figure 3. Based on the recommendation of this workflow, hundreds of shallow uphole (~100 m) and a few deeper uphole (~300 m) surveys were collected field-wide. A precise near-surface 3D model was then prepared using uphole and well data to compute statics caused
Figure 4A Stack with model and refraction statics applied.
Although an acceptable statics solution was achieved from the workflow discussed above, the identification of laterally varying high-velocity layers within the top 500 m of the shallow subsurface and their eventual mapping were critical to further refine the statics for a confident structural interpretation. This work is ongoing. However, a successful experiment to acquire high-resolution VSP (HRVSP) was tested, and is now routinely deployed in every exploration well to collect high-frequency VSP data. The purpose is to obtain the precise depth and velocity of the low- and high-velocity layers in order to construct an accurate near-surface model and calibrate seismic-derived statics. After several experiments and trial runs, HRVSP was acquired in 2 m intervals from 100 m to a total depth of 500 m below the surface, with a single vibrator placed at a 15 m offset from the wellhead sweeping from 6 to 120 Hz. A HRVSP dataset (up to ~110 Hz) was collected along with a full suite of shallow wireline logs. The match between the shallow synthetic, HRVSP, and the seismic data is excellent and the low- and high-velocity layers correlate very well (Figure 6). The results of this experiment were very encouraging to plan for HRVSPs and shallow logs at several locations and will be fully integrated in the seismic data processing.
Borehole data in seismic velocity picking Figure 4B Stack with model and refraction and reflection statics applied.
Figure 4C Stack with model and refraction and reflection statics and residual statics applied.
84
It is unfortunate but true that the most critical step in seismic data processing, velocity picking, is carried out without any consideration to the true geological velocity, which is readily available from borehole data, such as sonic log, check shots, and VSP. Arguments that the seismic velocities are picked to flatten the reflection are justified if the reflections are clearly visible and a quick qualitative check can guarantee a perfect flattening. The CDP shown in Figure 2 clearly shows that such a practice of velocity picking may not be valid, where even a 1% velocity change can provide very different stacks (Figure 7). It is therefore imperative to use borehole data to
© 2005 EAGE
first break volume
23, June 2005
special topic
Multi-disciplinary Geoscience
Figure 5 Comparison of final stack with first-breaks-based statics to a four-step statics workflow, which is a combination of refraction statics derived from the uphole data-based model, refraction statics using a layer-stripping approach, followed by reflection and auto statics. obtain interval velocities, which can be readily converted into stacking velocities, thereby providing an invaluable reference in velocity picking (Figure 8). This information should preferably come from VSPs, as they better c orrespond to seismic velocities due to the similar bandwidth of acquisition when compared with sonic log-derived velocities. As clearly shown in Figure 8, slight modifications in velocity picking matched to the borehole-derived velocities may not only pro-
Figure 6 Composite showing sonic velocities, HRVSP corridor stack and seismic stack correlation. The presence of several thin high-velocity layers is confirmed. A good correlation between the HRVSP corridor stack and seismic data may help refine the statics model.
© 2005 EAGE
vide better quality of stacks, but also may make seismicallyderived interval velocities more geologically meaningful.
VSP data in multiple identification As discussed, the presence of several shallow salt, anhydrite, and dolomite layers results in a series of primary and interbed multiples. Some of these multiples can be eliminated by cascading routine and specialized demultiple algorithms. However, others are difficult to remove without negatively
Figure 7 Stack sections showing 1% change vs stack quality, the data is very sensitive to the velocity picking.
85
special topic
first break volume
23, June 2005
Multi-disciplinary Geoscience
Figure 8 VSP can be an excellent source for providing reference stacking velocity trends in seismic velocity picking. This can be seen in panel 2 where the seismic velocity picking is matched with VSP-derived stacking velocity trend and in panel 4 where the interval velocity from the stacking velocity becomes more meaningful. affecting the primary events, due to insufficient moveout in both near and far offsets. This means that interpretation has to be carried out with remnants of multiples in the final data. In such circumstances, the best approach would be to somehow identify the multiples and exclude them from the interpretations. VSP, by virtue of its acquisition geometry, can be a very useful tool to help identify the primaries and respective multiples. The diagram in Figure 9 illustrates multiple generation in VSP. The multiple event (M1) terminates directly below where the strong impedance contrast of primary reflection (PR1) meets the first breaks at receiver depth level 4, as it cannot exist at receiver depths deeper than the event generating PR1. Both the time difference between primary and multiple arrivals, as well as the receiver depth level where a
Figure 10 VSP upwaves before and after deconvolution. Before deconvolution is applied it is possible to identify the multiple events, which are delayed in time but truncating at the same depth as the interface at which the multiple is reflected.
86
Figure 9 A diagram illustrating how the multiple event (M1) terminates well below the first-breaks at receiver level 4. The time-difference between primary and multiple arrivals and the receiver level where a multiple terminates provides information about the interval responsible for generating that multiple and also the interface that finally reflects that multiple. multiple terminates, provide information about the event responsible for generating the multiple. This information is critical for the success of pre- and poststack modelling-based multiple removal schemes. It is important to note that this interpretation can only be performed when both pre and post VSP downwave deconvolution upwave datasets are studied (Figures 10 & 11). Figure 12 illustrates the final results of borehole data integration efforts in seismic data processing, which led to confident structural and stratigraphic interpretation and several successful wells.
Conclusions The integration efforts undertaken have resulted in a high quality 3D seismic data set. Innovative use of upholes, log,
Figure 11 An integrated composite can be useful in understanding the correlation of primary events and also the possible identification of multiple events (as shown in the yellow box). The VSP corridor stack before deconvolution provides a good correlation of multiple events in the yellow box.
© 2005 EAGE
first break volume
special topic
23, June 2005
Multi-disciplinary Geoscience improvement. The HRVSP experiment designed and applied to a real case was the important part of the study, as it has the potential to further refine the statics and velocity model in the top 500 m of subsurface to achieve a precise statics computation and a confident seismic interpretation. In most seismic interpretation, zero-offset VSP data is only used in its fully-processed corridor-stack form. This information is sufficient for correlation purposes but, in our opinion, represents only partial use of the information available in the VSP wavefield. Analysis of VSP data at different stages of its processing can greatly aid seismic interpretation, especially of difficult seismic data, such as that from the Ledjmet blocks in Algeria.
Figure 12 2001 Vs. 2004 processing: The integration efforts undertaken resulted in a high-quality 3D seismic data. Innovative use of upholes, log and VSP data to improve the statics, velocities and primary-multiple identification were the key steps contributing to this improvement. and VSP data to improve the statics, ve locities, and primarymultiple identification were the key steps that resulted in this
Acknowledgements Special thanks to Cherif Hellal, chief geologist, Eastern Region, Sonatrach for his sustained contribution to the advancement of this project. Thanks to Todd Mojesky, Lavdosh Bubeqi, and Maria Bastidas for the technical support and processing of the seismic data. The authors also thank First Calgary Petroleums and CGG Canada Services for their approval to publish this work.
EAGE / SEG Research Workshop on
‘Multicomponent Seismic Past, Present and Future’ 5 – 8 September 2005, Pau, France This joint EAGE / SEG workshop will focus on all aspects related to current uses and future trends in Multicomponent Seismic Technology:
• • • • • • •
Survey design, spatial sampling, resolution and illumination Marine, land and borehole multicomponent acquisition systems 2D, 3D and 4D multi-component processing methods Lithology prediction, porosity estimation, fracture characterization, P-P versus P-S AVO Well calibration for P-P and P-S waves Multidisciplinary case studies that include and prove the value of converted waves Reservoir monitoring with S-S and P-S waves
WWW.EAGE.ORG
© 2005 EAGE
87
