Kaombo is a multifield development on Block 32 in Angola. Approximately two-thirds of the block is covered with a salt canopy that strongly degrades visibility, which prevented a good estimation of the resources and their distribution in several subsalt areas during the initial years of exploration. However, the last decade has seen a dramatic improvement in seismic processing, enabled by the increase in high-performance computing (HPC) capabilities. Through examples of subsalt fields of the Kaombo project, we show how new techniques have had a positive effect.
Introduction
In the Kaombo development, Tertiary turbiditic reservoirs are located below the salt canopy. Unlike other blocks of the lower Congo basin that are more proximal to the shore, a shallow canopy of complex-shaped salt bodies below the sea surface covers much of the block. The salt bodies create strong lateral seismic-velocity variations that are highly detrimental to classical time-processing approaches. Thus, prospect-maturation work during the early days of this block (1999 to 2001) was focused on targets outside of the salt influence.
While its application was initially restricted to a few areas for cost reasons, prestack depth migration (PSDM) rapidly became the standard exploration tool on the block during the period from 2002 to 2004. These first stages of subsalt explorations involved Kirchhoff and common azimuth wave-equation migrations, because of their relatively low computational cost, applied to a 3D narrow-azimuth data set. Migrations were performed by use of simple isotropic velocity models and a priori salt models where the salt geometry could not be assessed with certainty. The quality of the images was improved compared with standard time processing, but some work remained to increase the probability of success of subsalt prospects sufficiently to drill these prospects.
Application of depth-imaging techniques thus allowed a glance at the subsalt potential of the block. Depth imaging is a technique that takes advantage of modeling the propagation of waves in the space domain. Images are functions of data and a seismic-velocity model. In environments such as Block 32, complexity of the velocity model and uneven illumination of the subsurface by the data are severe challenges. The strategy of image improvement is therefore limited by three factors: the capability to accurately describe subsurface propagation parameters (velocity model), the capability to accurately model the wave propagation through the subsurface given the complexity level of the subsurface models (migration algorithms), and the capability to compensate for uneven illumination (acquisition). Progress in subsalt imaging can be achieved by improving methods in these three domains, which are unlocked by the increase of HPC capabilities and new acquisition designs.
Imaging for Exploration (2005 to 2009)
After initial successes of salt-free prospects on the block, intensive efforts were made to prepare and process data from a second narrow-azimuth acquisition over the entire block that was acquired in 2007. Several main improvements were made at this stage to obtain a better image. Upon acquisition of new data, two of these improvements were made on the velocity-model-building process, and the last one came from new imaging algorithms.
First, new structural concepts were derived from regional knowledge and previous images that led to a dramatic change in salt models. Initial images were not good enough to allow seeing the base of the salt in many areas of the block. Gravity modeling helped constrain the base-of-salt position in several areas, but, because of the heterogeneous nature of the sediments outside of the salt, inversions could not be relied on for accuracy. As a consequence, the base of salt was interpreted in many places after a structural model and not from direct observations of seismic images. In 2005, using existing PSDM data, a new structural model was built that led to a major reinterpretation of the salt on the basis of the estimation of the amount of shortening on the block and the observation of perched Cretaceous basins. This new model generally increased the thickness and complexity of salt bodies where the base could not be seen. Combined with the preparation of the new seismic survey, this triggered an intensive reprocessing effort in order to produce a consistent model over the entire block. This model was then used as a starting point for more-localized projects.
Second, anisotropy of seismic-wave-propagation velocity was accounted for by use of a tilted-transverse-isotropy assumption to process the newly acquired data. A general anisotropy 1D burial law was derived for the delta parameter. Although simplistic, this method led to a dramatic reduction of the well misties observed on previous isotropic PSDM images. Although misties remain, the general trend of these misfits with depth is removed, and most of the markers were tied with an error of less than 50 m, which was sufficient for exploration purposes.
Eventually, new imaging algorithms were used. Beam migrations were used because of their speed and the ability of some of their implementations to clean noisy data and give an image of very high dips. Shot-record wave-equation migrations (SRMs) also began to be widely used because their computation cost became affordable. Beam and SRM algorithms were therefore used together for their complementarity.
Imaging for Development (2010 to Present)
As wells were drilled and discoveries made, a development-hub perimeter was defined in the southwest part of the block. Named Kaombo, it comprises two subsalt fields. To prepare the development of these fields, a wide-azimuth (WAZ) streamer acquisition was designed to gain a better vision of steep dips below the salt that were observed in the wells but not on seismic images, and to obtain a better view of the reservoirs. Survey parameters were chosen after a careful modeling study to achieve these goals. Improving the data was the last axis of progress that was used because it is also the most expensive, but this WAZ acquisition provided a major quality step change because it illuminated areas that previously were unseen or weakly visible in the narrow-azimuth surveys (Fig. 1).
Another new migration algorithm, reverse-time migration (RTM), was also used to process this WAZ survey. RTM is an algorithm that models the full wavefields through time and not only its downward continuation. Its advantage over other wave-equation-based migrations is that it is able to produce an image of very highly dipping reflections, which makes it a preferred algorithm for complex environments such as subsalt. However, it is one order of magnitude more costly than SRM. Given that the computation cost of these wave-equation-based migrations is a function of the highest migrated frequency, the use of RTM has been restricted to a bandwidth that is narrower than the recorded data. Another drawback of the wave-equation-based methods is that production of common image gathers is traditionally not as straightforward as it is for asymptotic migrations: either the gathers are very costly or depend on a model to be reconstructed, or they do not represent an actual data-space parameter but rather different lags of a modified imaging condition in time or space. Common image gathers are a redundant view of the subsurface and therefore give the ability to apply additional processing to improve the image. For this specific WAZ processing, a new approach was used to calculate surface-offset gathers. This enabled the production of offset gathers in four different azimuth sectors. Because RTM is today the most accurate migration algorithm, surface-offset gathers from RTM have a better quality than classical Kirchhoff or beam-migration gathers.
Another novel technique was applied on the data set to improve the velocity model. Full-waveform inversion (FWI) was tested to update the velocity model. It locally helped to refine the salt interpretation in areas of poor visibility (Fig. 2), where the model resulting from inversion has a few differences with the initial salt model in an area where the salt interpretation is difficult. However, because of the absence of very low frequency in the recorded data, this technique was used only to improve the existing salt model; it cannot yet be used as a substitute for salt interpretation with usual marine-acquisition bandwidth.
These new techniques are increasingly costly in terms of computing effort, and their use is made possible by the exponential increase of computation capacity with time. This capacity increase is therefore offset by computation cost, and, in the end, the duration of depth-imaging projects is generally not much reduced by the growth in HPC facilities because result quality was given priority. Furthermore, the interpretation and quality-control steps remain very manpower-intensive. Eventually, development requires a reservoir model that is classically populated using attributes derived from seismic images. Obtaining these attributes requires an extended bandwidth that is very costly for wave-equation-based migrations. At present, the reservoir model is still built from a sedimentary model because these attributes cannot be clearly interpreted in the images with limited bandwidth and areas of insufficient visibility.
The Road Ahead: Drilling and Production
The combination of a powerful migration algorithm (RTM) and classical image-enhancement techniques on gathers produced an image that is suitable for reservoir-model building. The structures appear clearly and with enough resolution to lower uncertainties on correlations and well ties. However, even with better images, challenges remain for development. For instance, the position of subsalt structures in the seismic images depends on subsalt sediment velocities, which often cannot be determined without a higher-than-usual uncertainty. This results in not only vertical but also lateral shifts of the reflections in comparison with the actual position of reflectors. Because reservoir models are built from seismic data and the sparsity of exploration and appraisal wells does not necessarily constrain the geometry of the field well enough, drilling of development wells is expected to be challenging. This difficulty could be overcome partly by rapidly remigrating surface seismic while drilling after updating the velocity model from well information to tie the seismic to the well. Conversely, the well can be tied to the seismic while drilling by use of special well-seismic techniques.
This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper OTC 24132, “Kaombo, From Exploration Toward Development: A Decade of Progress in Subsalt Imaging,” by V. Martin, A. Douillard, L. Lemaistre, V. Clerget, and C. Gerea, Total E&P, prepared for the 2013 Offshore Technology Conference, Houston, 6–9 May. The paper has not been peer reviewed. Copyright 2013 Offshore Technology Conference. Reproduced by permission.