|IMPROVING STRUCTURAL IMAGING WITH BEAM PSDM - OFFSHORE NORWAY|
1ION GXT Imaging Solutions, 2Wintershall Norge ASA
In the mid-Norway offshore region under study in this exploration project, pernicious water bottom remnant multiples and other classes of scattered noise obscure low-reflectivity primary reflectors at target level, making it very difficult to establish the structural control on potential prospective reflectors. Diffracted and scattered multiples from both the sea-bed and overburden events are a particularly severe problem in this area, at times completely obscuring the deeper prospects.
A bespoke approach to de-noise and multiple suppression was undertaken, involving swell noise attenuation in conjunction with two passes of high-resolution Radon de-multiple, SRME, and finally a residual SRME approach.
Following the careful re-processing of the data, iterative Kirchhoff hybrid gridded tomographic inversion was employed to build a velocity model conformable with the observed structure and available well control. In order to maximally exploit the available signal energy, a controlled beam migration was also used to image steeply dipping fault plane and horizon reflectors. The ability of beam techniques to selectively remove energy not associated with the Fresnel zone of the output image space is particularly helpful in minimising high frequency and other noise. An interpretation based on both the Kirchhoff and beam preSDM image products has led to an enhanced understanding of the prospect geometries in this otherwise difficult area.
Whereas on the old vintage images, the BCU-reflector was barely visible, the new re-processing and imaging has made significant improvements to the clarity of the BCU and more importantly to underlying erosional remnant prospects, possibly of Jurassic age. Improved imaging of Permian and basement structures were also obtained, including clearer fault definitions.
In this case-study, we will outline the technologies used to achieve the enhancements, and compare results of Kirchhoff and Beam images together with the conventional vintage processing.
|DATA CONDITIONING AND PRE-PROCESSING|
The input data consisted of two legacy surveys which were taken through a processing sequence including zero-phasing, re-sampling, gain recovery, denoise, demultiple and Q compensation. Survey matching was performed to create a single dataset with consistent amplitude, phase and timing characteristics for input into velocity model building and subsequent imaging. As both input surveys had the same azimuth they were migrated together to create one post migration dataset. Prior to running the final migration both input surveys were regularised on their own natural acquisition geometries to ensure that artefacts due to the different acquisition geometries were minimised.
Significant effort was required to optimise the denoise and demultiple approach for these data. A total of four passes of noise attenuation were performed in CDP and common channel domains using a technique which decomposes data into frequency bands and identifies and attenuates anomalous amplitudes within affected bands according to user-specified time-variant thresholds. Additionally a radial trace filtering approach was used for linear noise.
It was recognised at the start of the project that the data suffered from significant multiple contamination and that this was going to be difficult to remove. Several different techniques were tested before the most successful combination of multiple passes of SRME and parabolic high resolution radon demultiple was adopted.
The steeply dipping aliased water bottom multiple noise in the data meant a 2 pass approach for the parabolic radon was necessary. The first pass applied a slow velocity trend NMO correction to the data before performing trace interpolation and applying radon demultiple. This was done to de-alias some of the steeply dipping water bottom multiple noise. The second pass of radon demultiple was then performed on the gathers NMO corrected with the picked velocity field.
3D SRME was tested on these data and it did a significantly better job at handling the diffracted multiples and ringing multiples beneath sea bed depressions, however at the target horizons the improvement was less significant and the cascaded 2D approach described was used instead.
|VELOCITY MODEL BUILDING AND IMAGING|
The PreSDM imaging involved 6 iterations of gridded tomography. The gridded methodology results in a depth interval velocity model where layer-generated artefacts are minimised, and has the considerable advantage of removing the interpretation element from the model building process. Typically, horizon picking is not required, as the process works to flatten every reflector rather than those at predetermined boundaries. The gridded method also removes the need to define a single gradient per layer, which can be inadequate to describe a complex velocity regime, such as that found within chalk. Instead instantaneous velocities can be calculated to flatten every reflector within certain constraints to produce a high-resolution velocity grid (Hardy, 2003).
The presence of sharp vertical velocity contrasts, for example, unconformities and chalk layers, has led to the development of a hybrid approach. If this is not used, the gridded tomography will tend to smooth the velocities through the boundary rather than inserting a sharp boundary into the model. The hybrid technique is an amalgam of the layered and gridded approaches, incorporating the best aspects of both schemes i.e. including the subtle variations calculated by the gridded tomography within sharp boundaries (Jones et al, 2007).
In the hybrid approach the data is divided into geological macro layers and the velocity updates are computed for one macro layer at a time. The initial velocity model is based on the first macro layer using a well derived gradient if available; otherwise a smooth function derived from the stacking velocities is used. The first set of gathers is then migrated for input to the auto-picking routine. A further refinement at this stage is to apply a mute based on the horizon at the base of the macro layer to prevent velocity “leakage” from the underlying layer. The auto-picking and tomography are iterated as required until a sufficiently accurate velocity model is obtained for the macro layer. The horizon at the base of the macro layer can then be picked and inserted as a layer boundary into the model, usually using a short offset stack migration for the interpretation. The model is now fixed at the layer boundary and subsequent iterations will only update velocities below this horizon. A new starting velocity, based on the next macro layer, is inserted at this stage prior to the next migration and the process is iterated through the remaining macro layers.
The initial preSDM was run using the final velocity model derived from the prestack time migration also performed as part of this project. This velocity field was close to correct, particularly in the southern parts of the survey area however the velocity field did not honour localised high velocity zones. The first iteration of the preSDM model updating attempted to address localised high velocity regions in the shallow at around 500-700m. Iteration 2 was run slightly deeper to a maximum depth of 2000m. This was to allow for a coherent event (Shetland) to be used as a control horizon, for better travel time computations to be estimated and therefore a better vertical velocity solution during the tomographic inversion process. Iteration 3 was a global velocity update to correct for the general background velocity to 5000m then iteration 4 included a step to apply a V0K velocity function on the flanks of the Nordland Ridge and into the Traena Basin (Cretaceous) to force the velocities where neither automatic or manual picking of CIP gathers was possible. The application of the V0K function was constrained using horizons provided by the interpreter. Iteration 5 included a similar approach in the Triassic.
Two full volume migration products were delivered as part of the project. Firstly an amplitude preserving Kirchhoff Pre-SDM was run and in addition it was decided to attempt to improve the deep imaging with a Beam migration.
Figure 1 below and Figure 2 compare the two migration results, the Beam migration clearly offers improved imaging of the deep structures.
Figure 2 Fast Beam preSDM
The major challenge in pre-processing was to obtain a satisfactory demultiple of the input data and through thorough testing of several techniques and a flexible approach this was achieved by cascading 2D SRME and high resolution parabolic radon demultiple. This followed by successful survey matching between the two input surveys ensured that we had gathers of a suitable quality for velocity model building and imaging
The final images from the Kirchhoff Pre-STM and Pre-SDM and the Beams migration show significant improvement over the vintage data. Significant improvements have been made to the clarity of the BCU and Permian reflectors as well as the basement structures and fault definitions.
The authors would like to thank Wintershall Norge ASA and their partners for permission to publish this work; Chris Tyson,
Hardy, P.B., 2003, High resolution tomographic MVA with automation, SEG/EAGE summer research workshop, Trieste.
Jones, I.F., Sugrue, M.J., Hardy, P.B., 2007, Hybrid Gridded Tomography. First Break, 25, No.4, p15-21.?