RESERVOIR MONITORING, 4D SIGNAL,

Transcription

1 RESERVOIR MONITORING, 4D SIGNAL, AND FIBER-OPTIC TECHNOLOGY Steve Maas, Rune Tenghamn, Brett Bunn, Natasha Hendrick, and Mazin Farouki (speaker) Petroleum Geo Services (PGS) DAY 1 SESSION 2 GEOPHYSICS PAPER 7 Introduction One of the challenges faced by reservoir engineers is to understand the way fluid saturation, pressure and compaction change between wells during production of a hydrocarbon reservoir. The optimum placement of infill wells and identification of new step-out opportunities to maximize recovery of hydrocarbons depends on such information. Time lapse (or 4D) seismic is acknowledged as being the only direct wide-scale reservoir management tool capable of revealing these important details about a producing reservoir. The Life of Field Seismic (LoFS) project at the Valhall field on the Norwegian continental shelf illustrates the technical and economic success of reservoir monitoring using 4D seismic data. Today, as the industry explores, drills and produces deeper and more challenging targets, reliable and consistent reservoir monitoring is becoming even more essential. A new fiber optic seafloor seismic acquisition system with high dynamic range, low background noise, low cost per channel and a long operational life is leading the way towards cost-effective permanent seismic reservoir monitoring. 4D Seismic for Reservoir Monitoring 4D seismic can be acquired using conventional 3D marine streamer acquisition, retrievable oceanbottom cables (OBC), seafloor nodes or permanent seismic installations. A benchmarking study of exploration and production operators conducted by PGS in 2007 (Rekdal, 2007) highlights the various factors to be considered when selecting a particular 4D seismic approach for monitoring and managing reservoirs. 4D seismic streamer surveys are used where shear-wave information is not required for optimum imaging of the reservoir, where infrastructure or weather won t negatively impact on the resolution of the 4D seismic data, and where repeat surveys are required only every year or so. Retrievable OBC surveys are considered appropriate when multi-component seismic data are important, but frequent surveys are not necessary. In the case that surface infrastructure impedes streamer surveys, or the seafloor is not suitable for OBC or trenching of permanent installations, node technology can deliver good quality multi-component 4D seismic data. However, since the ability of 4D seismic data to detect subtle changes in the reservoir is dependent on the repeatability and resolution of the data, the optimum technical solution for reservoir monitoring is permanent seafloor seismic installations. The permanent placement of sensors improves repeatability and the signal-tonoise ratio of the seismic data, enabling the detection of smaller 4D seismic responses. Permanent seafloor installations also make it possible to acquire shear-wave information, enable the acquisition of 4D seismic where surface infrastructure is significant, and are well-suited when frequent time-lapse surveys are required. The Valhall LoFS Project A review of the motivating factors behind BP s implementation of a permanent seafloor seismic acquisition system at the Valhall field, Norway, illustrates the positive impact 4D seismic can have on the recovery, costs and risks of a hydrocarbon production operation. The Valhall field consists of a structurally complex, weak chalk reservoir approximately 6km wide and 13km long, and up to 70m thick. The decision to implement a permanent seafloor seismic acquisition system at Valhall has been driven by a number of different issues (e.g. Barkved et al, 2004). First, pressure depletion of the Valhall chalk reservoir during production results in significant compaction of the reservoir, which in turn causes hardening of the reservoir. In contrast, the overburden tends to resist subsidence and stretches. The subsequent increase in seismic velocity in the reservoir zone and decrease in seismic velocity above the reservoir means 4D seismic is particularly well suited for delineating production-induced changes at 1 Geological Society of Malaysia

2 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 7 Valhall. Further, the structural complexity and low permeability of the reservoir means that a relative large number of wells are required to maximise hydrocarbon recovery. Thus, 4D seismic has an important role to play in the optimum positioning of new wells to avoid depleted zones and help prevent problems with penetrating pressure seals while drilling horizontal wells. The motivation for a seafloor 4D seismic operation at Valhall is also related to the difficulty in imaging the central part of the field using conventional P-wave data as a result of a shallow gas cloud. Shear-waves are largely unaffected by gas, so imaging of the entire Valhall field becomes possible with multi-component acquisition. In addition, the high repeatability of a permanently trenched 4D seismic system enhances the opportunity for the detection of small 4D effects. The LoFS project at Valhall commenced in 2003, with 4D monitor surveys initially collected every three months over more than 120 km of permanently trenched 4C seismic cables. Following each survey the P-wave and converted-wave (PS-wave) pre-stack depth migrated volumes, together with difference volumes and 4D attributes are derived to monitor production at Valhall. BP reports that fault patterns mapped from the converted-wave data have been essential in the planning and drilling of new production wells (Barkeved et al, 2004). In addition, acoustic impedance variations have successfully highlighted drainage patterns about horizontal wells, and relative travel-time shifts are mapping compaction effects both above and below the reservoir (Barkved and Kristiansen, 2005). The permanent installation allows excellent repeatability in the monitor surveys and demonstrates the robustness of the 4D method. Fiber-Optic Technology for Reservoir Monitoring At the time the Valhall LoFS project was implemented, only electrical permanent installations were available for seismic reservoir monitoring. While proven technology, conventional electrical seafloor seismic acquisition systems have bulky electronic components for signal detection and conditioning that can make cable deployment more problematic and relatively more expensive. Each receiver station on the seafloor must be powered, and the signal detected by each sensor is digitized and multiplexed at the receiver station before being transmitted back to the surface along electrical cables. The sensors and interconnecting cables of this electrical equipment are subject to leaks and corrosion, which shortens the life of the seismic installation. Thus electrical permanent systems are best suited for installations expected to operate less than 10 years. In contrast, the entire wet end of a fiber optic seismic acquisition system is passive there are no electronic components to fail as a result of water intrusion. Thus the life expectancy of a fiber optic system is expected to extend beyond 20 years. Elimination of copper wires from the cables make fiber optic seafloor seismic systems relatively lighter and easier to deploy, and elimination of high-voltage power makes them safer to handle in the field. Furthermore, use of standard telecom components in the construction of a fiber optic acquisition system contributes to relatively lower capital costs. It is apparent that when multicomponent 4D seismic data is required, passive fiber optic seismic acquisition systems can now deliver a more reliable and economically viable solution. Fiber-Optic Seismic Acquisition System design and field tests The new fiber optic seismic acquisition system presented here has been completely engineered from the ground up and has undergone multiple field trials over the past decade. The fiber optic sensors are optical transducers made using Michelson interferometers. The hydrophone is an air-backed mandrel wound with an optical fiber. The tri-axial accelerometer at each receiver station consists of three orthogonally mounted sensors in a pressure-balanced assembly. Incoming seismic energy causes a phase shift in the light that passes through these interferometers. The light is then transmitted back to the surface where the phase information is extracted to output a 32-bit digital signal equivalent to the incoming seismic wavefield. The key unique aspect of this new fiber optic seismic system is the Dense Wavelength Division Multiplexing telemetry scheme used to optically power the sensors. The ability to transmit unique light signals using bins only some eight-tenths of a nanometer wide implies many unique signals can be carried on a single optical fiber. Consequently the system can accommodate channel counts in excess of 2000 per cable this has a significant impact on reducing costs per channel. The wavelength division multiplexing is layered with frequency multiplexing this has several benefits over time-domain approaches including extra signal processing bandwidth yielding greater dynamic range (> 140dB). March 2009

3 DAY 1 SESSION 2 GEOPHYSICS PAPER 7 The opto-electronic surface instrumentation emits the frequency-modulated laser signal along the optical fibers and processes the returned optical signals to extract the desired seismic signal. This active and expensive equipment including laser sources, demodulators and the electronic acquisition components is kept at the dry end of the system, making for easy maintenance and updates, and extending the life expectancy of the permanent seismic installation. A 2D seismic field trial has been conducted in shallow water in the Gulf of Mexico. The new fiber optic acquisition system was deployed next to a conventional electrical seafloor cable, and a preliminary examination of the data has been completed. Figure 1 compares hydrophone data acquired by co-located sensors from these two systems. The fiber optic system has accurately recorded the seismic pressure wavefield. Seismic reflection events which appear as laterally coherent signal exhibiting hyperbolic moveout on the common-receiver records in Figure 1(a) are consistently detected by both acquisition systems. In addition, the frequency-domain representation of the two datasets is very similar. The same holds for the 3-component velocity data recorded by the two acquisition systems. A useful tool for analysing vector fidelity of a multi-component seismic sensor is hodogram analysis. A hodogram is a plot of seismic amplitude over time; a display of the particle-motion path for the seismic energy. A plane wave measured by a three-axis sensor should be linearly polarized. Figure 2 shows the hodogram analysis for the direct P-wave arrival recorded by a representative fiber optic receiver for another seismic field trial conducted in the North Sea. The linear particle motion and accurate polarization of the seismic wavefield recorded by the fiber optic sensor illustrates the excellent vector fidelity of the sensor. Conclusions While measuring pressure and fluid properties at a well is essential for validating a reservoir s performance, only 4D seismic can help operators understand and define characteristics of the reservoir away from the well. The robustness of the 4D seismic method is well illustrated at Valhall, where a permanent seismic seafloor installation has been used by BP since 2003 to successfully monitor changes in reservoir compaction, identify bypassed oil, and optimize the location of new wells. A new fibre optic multi-component seismic acquisition system for permanent installation on the seafloor has been designed and tested. The ability of fiber optic technology to deliver significant improvements in signal quality, acquisition consistency and long-term reliability that cannot be achieved with conventional electrical seafloor installations offers the opportunity and motivation for the oil and gas industry to accelerate the implementation of permanent seismic reservoir monitoring and reap the financial benefits of incorporating 4D seismic in to the reservoir management process. Acknowledgements The authors would like to thank BP and Valhall partners for permission to use the Valhall examples. The authors would also like to thank PGS for permission to publish this work, and the many scientists and engineers within PGS who have contributed to the design, construction and utilisation of the new fibre optic seafloor seismic acquisition system. References Barkved, O.I., Kommedal, J.H., and Thomsen, L., The role of multi-component seismic data in developing the Valhall Field, Norway: EAGE Extended Abstracts, Paris, E040. Barkved, O. I., and Kristiansen, Tron. Seismic time-lapse effects and stress changes: Examples from a compacting reservoir: The Leading Edge, December Rekdal, T., Fiber-optic reservoir imaging: E&P, September Geological Society of Malaysia

4 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 7 Figure 1: Hydrophone data acquired by a conventional electrical seafloor seismic system and the new fibre optic seafloor seismic system: (a) a representative common-receiver gather; and (b) average frequency amplitude spectra of the data shown in Figure 1(a). An 8/10-80/120Hz bandpass filter has been used to display the seismic data. The fibre optic system is accurately detecting the seismic pressure wavefield. Figure 2: Hodogram analysis on the direct P-wave arrival recorded by a representative multicomponent fibre optic sensor with a shot-receiver azimuth of 45 degrees. Left: inline and crossline velocity traces with the direct P-wave window indicated; and right: particle-motion hodogram for the direct P-wave arrival. The linear particle motion and accurate polarisation demonstrate the excellent vector fidelity of the fibre optic sensor. March 2009

5 DAY 1 SESSION 3 GEOPHYSICS PAPER 9 BEAM DEPTH MIGRATION FOR IMAGING OF COMPLEX GEOLOGY Karl Schleicher 1, John Sherwood 1, Lynn Comeaux 2 and Mazin Farouki 2 (speaker and corresponding author) 1 Applied Geophysical Service, AGS, Houston, which has now been acquired by PGS. 2 PGS, Kuala Lumpur. Introduction Kirchhoff migration has traditionally been the leading implementation for application of depth migration to seismic data. There are many reasons for this, such as efficiency, ability to image steep and even overhanging dips, and flexibility. In most parts of the world Kirchhoff migration produces images that are as good as, or better, than the more expensive implementations using downward continuation algorithms. However, the limitations of Kirchhoff migration are well known and its inability to image more than a single arrival is the most damaging. Downward continuation algorithms, on the other hand, handle all arrivals but their inability to image steep dips is a severe limitation. In geological situations where there is a complex overburden and the signal to noise ratio in the regions that are the target for exploration are low, Kirchhoff migration often fails to produce good images and if the target dips are steep, downward continuation algorithms cannot be used as an alternative. Such situations occur for example where complex and rugose salt bodies mean that very little energy reaches the target, where basalt layers stop most of the energy from reaching the target, or where the targets are faults or fractures in the basement. In these situations, the single arrival imaging of Kirchhoff migration fails to give a good image. Instead, artifacts caused by the swinging action of the migration often obscure the real targets and it is very difficult to distinguish artifacts from geology. Summary Beam Migration provides an effective alternative to Kirchhoff or wave equation migration. Beam Migration relaxes the single arrival limitation of Kirchhoff while retaining its steep dip capability. There are a number of different types of Beam Migration implementations such as Hill s Gaussian Beam Migration. Our implementation of Beam Migration is unique in the industry and involves a decomposition of the data into dip components using the Radon transform and a back-propagation of the dip components into the earth. The dip components can be enhanced based on various criteria before the back-propagation, thereby giving a more coherent image. The methodology inherently allows the attenuation of multiple energy, and coherent as well as non coherent noise. Our implementation of Beam Migration (which we refer to as BPSDM for Beam Pre-Stack Depth Migration) has merits of simplicity, economy, flexibility, and future development possibilities. Migrated images have excellent accuracy and quality, especially in areas of poor signal to noise ratio and steep dip. The relative economy makes it an excellent velocity estimation tool to use prior to other, more computation intensive, depth migration methods. Method The input seismic data is assumed to have been prepared with conventional preprocessing. Thereafter BPSDM consists of three important steps, decomposition, migration, and reconstruction. 1. Decomposition A multidimensional slant stack decomposes the data into seismic wavelets, from msecs. in time duration, within local surface spatial superbins each being centered on a uniform Cartesian grid of CDP 5 Geological Society of Malaysia

6 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 9 xline and inline co-ordinates (x,y) and S-R half offset co-ordinates (hx,hy). Each wavelet should thus have a center location (x,y,hx,hy,t) and determined dip components (dt/dx, dt/dy, dt/dhx, dt/dhy), these contributing to the properties of the wavelet. For each spatial axis the grid interval between superbin centers is typically around 200 to 400 meters, and the superbin width is chosen somewhat larger in order to have overlap. It is significant to note that this decomposition performs a desirable uniform spatial gridding or binning of the data, the result being essentially independent of minor variations in the data acquisition geometry (minimized acquisition footprint). Also the individual wavelets should not exhibit any aliasing effects, even for very steep time dips. Thus the BPSDM procedure bypasses the aliasing issues that are a very significant problem in both Kirchhoff and downward continuation wave equation migration. The wavelet amplitude is preserved, a vital issue for any future AVO or inversion operations, and some random noise suppression is achieved at this early stage. 2. Migration Given a wavelet s center (x,y,hx,hy,t) and dip components (dt/dx, dt/dy, dt/dhx, dt/dhy ), plus a current earth velocity model, it is possible to 3D ray trace from source and receiver locations and determine some corresponding best reflector migration location (xmig,ymig,zmig), together with ancillary properties such as reflector dip, reflector azimuth, angle of incidence at reflector, S & R wave front curvatures, local interval velocity, etc. It is of paramount significance that a point to point mapping exists between the unmigrated and the migrated center of each seismic wavelet. This means that the seismic wavelet decomposition forms a basis in the migrated domain as well as the unmigrated domain, the mapping function being the earth velocity model. Note that migration is applied correctly to each coherent wavelet and this bypasses the multipath traveltime problem encountered with Kirchhoff migration. Also note that BPSDM does not suffer from the severe steep dip and turning wave limitations of typical wave equation migrations. Economics often force aperture limits in Kirchhoff migration which can result in not imaging steep dips and turning waves. The BPSDM independent point to point mapping of each wavelet has no migration aperture issue. Since S & R rays corresponding to a wavelet s properties normally will not intersect to give a model two way traveltime exactly equal to the wavelet center time t, it is also sensible to estimate a focussing quality factor, q. This is an additional valuable wavelet property. If the wavelet is truly a primary reflection, then q is representative of traveltime discrepancy along the composite S & R ray path and can be included in a tomography method. Alternatively, a poor focus value can be used to recognize a multiple reflection wavelet, or a converted wave event, with its NMO offset dip (dt/dhx, dt/dhy) significantly different from a corresponding primary reflection. In summary, for a current earth velocity model, the migration operation determines and stores reflector location and associated properties along with each wavelet and its surface location and dip parameters. This stored point to point mapping between unmigrated and migrated space is very valuable and is not directly available with the Kirchhoff and wave equation methods. 3. Reconstruction A seismic wavelet can be easily contributed to its local region in either unmigrated space or migrated space. In particular, the local nature of a wavelet and its associated migration properties enables a very limited wavefront Kirchhoff migration contribution to a 3D migrated depth volume (x,y,z) for the common offset (hx,hy). This yields certain improved signal to noise characteristics over normal full wavefront Kirchhoff migration, where data from millions of seismic traces do not necessarily cancel in an output quiet reflector area, such as salt. Relevant ray path spreading properties enable amplitude correction of each primary reflection for its actual propagation path through the interval velocity model. This facilitates later AVO or inversion operations. Note that any wavelet can be excluded or weighted down based upon a variety of individual or joint criteria for the wavelet properties, for example the quality of focus, thereby providing powerful flexibility for coherent noise reduction in the unmigrated or the depth migrated data. Inline, xline or full volumes are output March 2009

7 DAY 1 SESSION 3 GEOPHYSICS PAPER 9 on appropriate grids, both for quality control and for residual moveout analysis on common reflection point depth gathers. The residual moveout field is interpreted either manually or automatically, depending on the complexity of the data. This information is supplied to a tomography routine and enables the updating of the earth interval velocity model. The migration, reconstruction, and tomography steps are then iterated until a satisfactory velocity model is developed for which the common reflection point depth gathers are adequately flat. The final common offset volumes are reconstructed on an appropriately fine (x,y,z) grid. Since this operation is reasonably economic, it is normal to output volumes over the entire (x,y) common mid-point range of the input data. Results Results from the Beam migration are proving to be superior to Kirchhoff and wave equation results in several respects. Examples are shown in Figures 1 and 2 below. The quality and flexibility of the AGS implementation of BPSDM has been illustrated in many ways. These include its unique capability for handling steep and overturned dip, its demultiple options, the ability to handle extraneous coherent noise, and the adaptability to anisotropic velocity earth models. Also of importance is its speed for model iteration and the calculation of residual 3D RNMO for input to 3D tomography, the extension to multi- and wide azimuth data acquisition, and the capacity to handle both land and marine data. Figure 1 is a comparison of Kirchhoff pre-stack time migration and BPSDM applied to data from the Beaufort Sea (both plotted in time to help direct comparison). BPSDM was very successful at imaging steep dips where Kirchhoff PSTM had failed. Based on these images the main structure to the left of the section which had been interpreted as a shale diapir in the time migration is re-interpreted as an inversion anticline in the BPSDM. Reservoir rock was identified in the core of the fold. The prospect was drilled and, in addition to confirming the re-interpretation, was awarded the single most significant discovery in the Beaufort Sea. Figure 2 shows CRP offset gathers and stacks created using BPSDM on data from the KG-D4 basin off the east coast of India. After several iterations of migration and tomography there was still residual moveout due to unresolved local velocity anomalies associated with a rugose water bottom, slump sequences, and a strong velocity increase at the basement. After our efforts to solve the problem using velocity estimation and depth migration, further improvement was achieved using a non-hyperbolic residual moveout. The migrated offset gathers output by BPSDM allowed the flexibility required to address these data problems. Conclusions BPSDM does have the anticipated merits of simplicity, economy, flexibility, and future development possibilities. Migrated images have excellent accuracy and quality, especially in areas of poor signal to noise ratio and steep dip. BPSDM s relative economy makes it an excellent velocity estimation tool to use prior to other, more compute intensive, depth migration methods. Acknowledgements The authors thank Devon Canada for permission to show the Beaufort Sea data, and Reliance for permission to show the KG-D4 data. We also thank the AGS data processing and interpretation staff for their efforts applying BPSDM to numerous datasets. References Rieber, F., 1936, Visual presentation of elastic wave patterns under various structural conditions: Geophysics, 1, an excellent velocity estimation tool to use prior to other, more compute intensive, depth migration methods. 7 Geological Society of Malaysia

8 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 9 Figure 1: Kirchhoff prestack time migration (left) and BPSDM (right) on Beaufort Sea data. The beam migration imaged steep and overturned events. Figure 2: BPSDM from the KG-D4 basin, East of India. The stack and migrated gathers in the top panels show residual moveout due to local unresolved velocity anomalies associated with a rugose water bottom and slump sequences. Residual moveout can be used to flatten the gathers and improve the stack, shown in the bottom panels. March 2009

9 DAY 2 SESSION 4 GEOPHYSICS PAPER 11 MITIGATION OF DRILLING RISK USING CONTROLLED SOURCE ELECTROMAGNETIC SURVEYS: CSEM WORKFLOW AND CASE STUDY Lars Lorenz 1, A. Muralikrishna 2, Anil Kumar Tyagi 2, Rabi Bastia 2, and Hans E. F. Amundsen 3 1 Reliance Industries Ltd., Mumbai 2 EMGS Asia Pacific, Kuala Lumpur 3 EPX AS, Oslo, Norway l lo@emgs.com Figure 1: Overview map, showing the position of the survey area at the East Coast of India. The targets, discovery Alpha and prospect Beta are depicted. SUMMARY Remote sensing of hydrocarbon reservoirs using controlled source electromagnetic (CSEM) surveys is a powerful tool in de-risking exploration prospects. This study illustrates the general workflow for the planning, execution and interpretation of seabed logging (SBL) surveys, and shows how the data set is utilized by Reliance in the selection of drilling targets in the East Coast Deep Water blocks, resulting in a significant gas discovery. INTRODUCTION The Krishna Godavari Offshore Basin off the East Coast of India is a highly interesting area for explorationists (Figure 1). Several world class oil and gas discoveries have been made in this basin. Reliance had a large gas discovery off the east coast of India in 2002 which came on-stream during late The basin comprises a wide range of depositional settings from coastal plains, deltas, shelf-slope aprons to deep sea fans. Commercial accumulations of hydrocarbons occur in sediments from the Permian to Pliocene. The main offshore hydrocarbon potential lies in the tertiary channel levee overbank Miocene to Pliocene reservoirs in the deep waters as the currently targeted play type. Due to the large amount of acreage owned by Reliance and the significant costs associated with drilling in water depths of 200 m to more than 2000 m, SBL was chosen as one of the methods to mitigate the drilling risk. METHOD The SBL technique is described in detail by Eidesmo et al. (2002) and Ellingsrud et al. (2002). A horizontal electrical dipole (HED) emits a low frequency electromagnetic (EM) signal into the seabed and downwards into the underlying sediments. EM energy is rapidly attenuated in the seafloor sediments due to low resistivity saline pore fluids. In high-resistive layers such as hydrocarbon filled sandstones, and at a critical angle of incidence the energy is guided along the layers and attenuated less (Kong et al. 2002). Energy is constantly refracted back to the seafloor and is detected by seafloor EM receivers. When the source-receiver distance (offset) is comparable to or greater than twice the depth of reservoir burial, the refracted energy from the resistive layer will dominate directly transmitted energy (Johansen et al., 2007). The detection of this guided and refracted energy is the basis of SBL (Ellingsrud et al. 2001). 9 Geological Society of Malaysia

10 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 11 Figure 2: 3D feasibility model (bottom) and the NMvO response (top) for a scenario with gashydrates and without gas hydrates. PRE-SURVEY STUDY Many wells drilled to this point in the channellevee complex penetrated a thick Pliocene-Pleistocene section with reservoir sand thickness varying from a few millimeters to 60 m thick sands over a gross interval of about 350 m. Depending on the thickness of the beds, the resistivity of the reservoirs varied between 30 ohmm and 1 ohmm. Shale resistivities were around ohmm, and water-bearing sand formations ohmm. These low reservoir and background resistivities pose challenges for CSEM due to low resistivity contrasts and high attenuation. Therefore, a careful evaluation of the feasibility of the survey had to be performed. The first step in a feasibility evaluation is 1D modeling. Even though strong simplifications for the geometry are introduced, the fast computation time makes it a powerful tool for parameter evaluation like frequency, background resistivity, reservoir resistivity and reservoir thickness. 1D feasibility showed magnitude responses of more than 2.0 for all but the most pessimistic scenarios and lowest frequencies. This encouraged proceeding with 3D modeling, using a time- domain finite difference code (Maaoe 2007). For the 3D modeling, the reservoir thickness was scaled to the grid cell size and the resistivity adjusted accordingly, keeping the thickness resistivity product constant. To evaluate the influence of the presence of gas hydrates, 3D modeling with and without the gas hydrate layer was performed. Even though the introduction of the gas hydrate layer challenged the ability to delineate the inner boundaries of the targets, the outer boundaries were still well defined and satisfactory responses for the discovery as well as the prospect were given (Figure 2). Based on the pre-survey 3D modeling study, the line was positioned along the long axis of both targets as this configuration proved to give the highest response, as well as being the most economical acquisition strategy. A receiver spacing of 1.25 km was chosen to ensure sufficient data coverage for more advanced data processing methods such as depth imaging or inversion. In addition, pre-survey studies showed a low sensitivity to the most common base frequency of 0.25 Hz, as well as a limited offset range above the anticipated noise floor for frequencies beyond 1.25 Hz. This resulted in the decision to create a composite square pulse as source signature with the highest energy distribution on as many frequencies as possible between these two boundaries, using the approach described by Mittet and Schaug-Pettersen (2007). SURVEY RESULTS Figure 3 shows the normalized magnitude response for three different frequencies. A steady increase in the overall responses occurred from NW to Figure 3: Normalized Magnitude (top) with steady increase in resistivity from NW to SE and local anomaly towards SE. Phase split depth conversion results plotted versus seismic (bottom) with consistent depth estimate over prospect. March 2009

11 DAY 2 SESSION 4 GEOPHYSICS PAPER 11 SE along the line both within, as well as outside, the mapped prospects. A defined anomaly was positioned over the south-eastern prospect while the response in the north-western discovery area was dominated by the steady increase in the overall response. The change in response for the two areas outside of the prospect areas towards NW and SE indicates a change in background response, either in the form of a gradual change or an abrupt increase. As the measurements towards SE of the prospect area show minor differences in response, an assumed gradual change has to be defined by a small gradient. Based on this observation, it was concluded that a reference towards NW is more representative for the NW part of the line and a reference towards SE for the SE part of the line. These different reference receivers were then used for a first-pass depth conversion by utilizing phasesplit analysis as described by Johansen et al. (2007) and the depth estimates superimposed on the seismic (Figure 3). The depth estimates for the response were consistent in the area of the south-eastern prospect and coincided well with the seismic anomaly. In the area of the discovery, a larger scatter in the depth estimates was observed, which was in line with expectations, as the strong 3D geometry of the target strongly violates the 1D assumption behind this simple depth conversion method. Figure 4: 3D inversion results versus inline systematics and phase-split depth conversion. 3D INVERSION AND POST- INVERSION MODELING Unconstrained 3D inversion of the data was performed after the fast-track data assessment indicated the presence of a defined anomaly. Different starting models were used to ensure that the result was not dependent on the starting point. Consistent inversion results were obtained and the imaged resistor is conforming to the standard attribute analysis for the lateral extension (Figure 4). No evidence for the known discovery is present in the 3D inversion result. Following the unconstrained 3D inversion, post inversion modeling was performed to include the additionally available geo-information in the form of seismic horizons and available well logs. Two different models were developed to explain the resistivity change outside of the discovery and prospect area between NW and SE. The first model is characterized by a gradual change of resistivity with water depth while the second model utilizes the knowledge about the existence of gas hydrates. The gas hydrate thickness is increased for this model according to the increase of the gas hydrate stability window (Figure 5). A good data fit is obtained for the reference areas for both models. Introducing a high- resistive layer for prospect Beta improved the data fit for both models as well. At the other hand, the model with gradual resistivity change shows an increased Figure 5: 3D model with gradual resistivity change (left) and gas hydrate thickness increase (right) from NW to SE. 11 Geological Society of Malaysia

12 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 11 data misfit when including the discovery Alpha, while an improved data fit occurs for the gas hydrate model. The second model opened up the possibility of having an anti-model to a thin resistor at prospect level. Considering the existence of gas hydrates, it was necessary to evaluate if higher resistive gas hydrates in the prospect area could result in a comparable response as for the prospect. Evaluation of this scenario showed that the magnitude gave a comparable response as a higher resistive prospect, but a very different phase response, which did not explain the acquired data at all. In addition, the gas hydrate resistivity of 5 ohmm used was considered at the upper limit for a continuous gas hydrate layer. This led to the conclusion that the anti-model was improbable. Integration Discussion The defined CSEM anomaly correlated spatially well with the outlines of prospect Beta. Phase- split depth conversion as well as 3D inversion imaged the resistor at the depth interval for the seismic prospect. The implications of the measured background resistivity change were evaluated by deriving two different scenarios. While either of the scenarios required the presence of a strong resistor in the lateral and vertical position of prospect Beta, one of the scenarios allowed explanation of the data without including the discovery. Preference was given to the scenario which allowed for the introduction of discovery Alpha. Based on this scenario, an anti-model was developed. The anti-model could not explain magnitude and phase data together, highlighting the importance of combined phase and magnitude interpretation. In addition, the required parameters were not considered by the Reliance asset team to conform to the anticipated behavior of gas hydrates in the area. Based on this evaluation, the drilling decision was made, and the presence of a commercial hydrocarbon accumulation at target level was proven (Figure 6). References: Eidesmo, T., Ellingsrud, S., MacGregor, L.M., Constable, S., Sinha, M.C., Johansen, S., Kong, F.N. & Westerdahl, H. [2002] Sea Bed Logging (SBL), a new method for remote and direct identification of hydrocarbon filled layers in deepwater areas, First Break, 20, March, Ellingsrud, S., Sinha, M.C., Constable, S., MacGregor, L.M., Eidesmo, T. & Johansen, S. [2002] Remote sensing of hydrocarbon layers by Sea Bed Logging (SBL): results from a cruise offshore Angola, The Leading Edge, 21, Johansen, S.E., Amundsen, H.E.F and Wicklund, T.A. (2007) Interpretation example of marine CSEM data. The Leading Edge, Volume 26, Issue 3, pp Kong, F.N., Westerdahl, H., Ellingsrud, S., Eidesmo, T. & Johansen, S. (2002) Seabed logging: A possible direct hydrocarbon indicator for deepsea prospects using EM energy. Oil & Gas Journal, May 13 Mittet, R. and Schaug-Pettersen, T. (2007) Shaping optimal transmitter waveforms for marine CSEM surveys, SEG Technical Program Expanded Abstracts 2007, pp Maaø, F.A. (2007) Fast finite-difference time-domain modeling for marine-subsurface electromagnetic problems, Geophysics 72, pp. A19-A23 Zach, J.J., Bjoerke, A.K., Stoeren, T., Maaoe, F. (2008) 3D inversion of marine CSEM data using a fast finitedifference time-domain forward code and approximate Hessian-based optimization, SEG Technical Program Expanded Abstracts 2008 March 2009

13 DAY 2 SESSION 4 GEOPHYSICS PAPER 12 MARINE MAGNETOTELLURIC (MMT) MAPPING OF BASEMENT AND SALT BODIES IN THE SANTOS BASIN, BRAZIL Sergio L. Fontes 1, P. de Lugao 2, Max A. Meju 3, V.R. Pinto 1, E.U. Ulugergerli 4, E.F. La Terra 1, L.A. Gallardo 5 1 Observatório Nacional, Rio de Janeiro, Brazil. sergio@on.br 2 StrataImage Ltd, Rio de Janeiro, Brazil. patricia.lugao@strataimage.com 3 Petronas Research Sdn Bhd, Subsurface Technology, Malaysia. maxwell_meju@petronas.com.my 4 Canakkale Onsekiz Mart University, Turkey. emin@comu.edu.tr 5 CICESE, Ensenada, Mexico. lgallard@cicese.mx Key Words: Electromagnetic imaging, marine magnetotellurics, basement and subsalt mapping. The marine magnetotelluric (MMT) method is rapidly emerging as a practical electromagnetic tool for investigating the deep resistivity distribution beneath the sea-floor and aid the exploration for hydrocarbons especially in areas of poor-seismic data. Remarkable advances in field instrumentation over the last few years now permit the use of MMT for hydrocarbon exploration in the marine environment (Constable et al., 1998; Sandberg et al., 2008). The MMT method has been successfully used to achieve marine exploration objectives such as imaging sub-basalts, carbonates and subsalts in situations where seismic imaging is poor. The high contrast in electrical resistivity between salt bodies and the surrounding sediments makes for a good target and provides an opportunity to test the applicability of the MMT method in oil exploration studies. The deep water basins in offshore Brazil are the sites of giant oil and gas discoveries, and should provide excellent test sites for evaluating the MMT method. Recent subsalt oil discoveries in the deep waters of the Santos basin in Brazil have received considerable attention due to the extent of the resource reserves and the challenges in exploration and production in such environment. This paper presents the first large-scale MMT survey that was acquired recently in Brazil by WesternGeco Electromagnetics in cooperation with Observatório Nacional/MCT and Petrobras as part of a major project on improving depth imaging by integrating multiple geophysical measurements in the Santos Basin. The regional gravity anomaly map shows a clear NE-SW structural trend in the area of study. 2D inversion of the MMT data across this trend would seem to be satisfactory. The 2D inversion model clearly shows resistivity anomalies that correctly delimit the position and depth extent of the known salt bodies. Our result thus provides an important practical validation of the MMT method in the deep marine environment of Santos basin. This study has demonstrated the feasibility of economically obtaining a reliable image of subsurface resistivity in the depth range of current interest in hydrocarbon development in offshore Brazil. MMT data fidelity processing procedures were developed that so far yielded high quality data. 13 Geological Society of Malaysia

14 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 13 NEAR SURFACE RESISTIVITY RESPONSES TO LITHOSTRATIGRAPHY AND FLUID CONTENTS Zuhar Tuan Harith 1, Ani Aiza Ashaari 1, Rosli Saad 2 1 Geoscience & Petroleum Engineering Dept. Universiti Teknologi PETRONAS, Bandar Seri Iskandar Tronoh. Perak Darul Redzuan. Malaysia 2 Geophysics Group,School of Physics, University of Science Malaysia, USM Penang. Malaysia 1 zuharza@petronas.com.my Resistivity variation is an important physical property to investigate near surface structures such as salt water intrusion, sinkholes, cavities, as well as active faults. The electrical property of the rocks is severely affected by pore connectivity and fluid type, thus their subsurface variation can be an indicator of the subsurface geology and its fluids. Shallow fractures, faults, and unconformities are expected to have low resistivity as they normally behave as conduits for ground water flow. However when those fractures and faults were filled with resistive fluid (such as liquid hydrocarbon and gas), it should reads a high resistivity. A electrical survey was conducted at two sites in Labuan Island to investigate the resistivity response of different lithology and fluid properties (FIGURE 1). Site 1 is an abandon coal mine where coal has been extracted for commercial used during the colonial period between 1847 and The tunnels in this site were filled with water. The second site is known as Bukit Minyak. On this site, the Belait (Late Miocene) and Temburung (Early to Late Miocene) formations are exposed. The Belait Formation is a lacustrine to shallow marine clastic sequence characterized by a thick conglomeratic bed at its base and the presence of coal-seams / layers. The Temburung on the other hand, is a deep marine clastic sequence. The unconformity contact between the Belait Formation and the Temburung Formation is very well defined. The sandy layers above and below the Belait-Temburung unconformity contact have a distinctive aromatic hydrocarbon smell indicating the presence of hydrocarbon seepage. In one locality an active seep can be seen. A schematic log section at the Site 2 is given in FIGURE 2. The equivalent of the Belait Formation offshore of Sabah is a proven major oil reservoir. Coal usually has a high resistivity compared to other sedimentary rock types. However when it is highly saturated, the layer could acting as a good conductor, thus has a low resistivity. An acidic mine water, conductivity could approach about 5000microS/cm, which corresponds to a resistivity of 2 ohm-meters (Johnson, 2003). The resistivity results were compared with the surface observation records. FIGURE 3 shows a resistivity image of from Site 1. The image clearly shows two distinctive northwest dipping low resistivity features indicating the presence of two coal beds. The upper coal layer was mine and the voids / tunnels are filled with water. The second coal bed is slightly deeper and probably un-mined. FIGURE 4 is a resistivity image of from Site 2. In general, the resistivity profile also shows a northwest dipping features, which is in concordance with general dipping pattern of the area. The exposed 3m coal layer is easily recognized as high resistivity zone (dry coal), whereas the conglomerate showed a low resistivity zone. Underneath the conglomerate zone, lies a high resistivity zone. This zone corresponds to the unconformable contact between the Belait and Temburung formations. The high resistivity zone indicates that the fluid flowing through this unconformity is a resistive fluid. Oil seepages flowing through a Belait- Temburung contact found 75 m to the southwest of our line, suggested that the resistive fluid in the study area is probably a mixture of oil and water. The shallow accumulation may be due to the migrated hydrocarbon through an inclined unconformity from the deep seated accumulation, within the petroliferous Belait and Temburung formations. REFERENCES JOHNSON, W. J., Applications of the electrical resistivity method for detection of underground mine workings. Geophysical Technologies for detecting underground coal mine voids, Lexington, KY, July March 2009

15 DAY 2 SESSION 4 GEOPHYSICS PAPER 13 Figure 1: A Google image of Site 2. A well defined Belait and Temburung unconformity contact can be observed here. Inside is a simplified location map of study area. Figure 2: Schematic log section at Site Geological Society of Malaysia

16 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 13 Figure 3: Resistivity image of Site 1, showing two distinctive coal beds. Figure 4: Resistivity image of the Site 2, showing high resistivity coal region and possible shallow HC accumulation. March 2009

17 DAY 2 SESSION 4 GEOPHYSICS PAPER 14 THE EFFECT OF RESISTIVITY ANISOTROPY ON EARTH IMPULSE RESPONSES Folke Engelmark 1, Bruce Hobbs 2 and Dieter Werthmüller 3 1 folke.engelmark@pgs.com Petroleum Geo-Services 2 bruce.hobbs@pgs.com 3 dieterwe@student.ethz.ch INTRODUCTION Resistivity anisotropy arises through a variety of scales from micro (e.g. grain size, pore water connectivity) to macro (e.g. laminated sand-shale sequences). For general anisotropy the physical property under consideration may vary in all three spatial directions. The simplest problems involve transverse anisotropy where resistivity at a point in any direction in a plane differs from the value perpendicular to the plane. We are here concerned solely with transverse anisotropy with a vertical axis of symmetry (TIV) so that resistivity at a point has a constant magnitude in any horizontal direction. Induction logs, laterolog and LWD (logging-while-drilling), at least in vertical wells, may be used to examine TIV in particular and these well log results often differ from indirect determinations of resistivity through DC resistivity and general EM surveying. Much of the earlier EM literature considered resistivity as isotropic but there is now great emphasis on the inclusion of anisotropy in modeling and inversion studies. In this paper we consider the effects of transverse anisotropy (specifically TIV) on the earth s electromagnetic impulse and step responses. THE MULTI-TRANSIENT ELECTROMAGNETIC METHOD In the multi-transient electromagnetic method (Ziolkowski et.al, 2007) current is injected into the ground between two electrodes (the source) and the resulting potential difference is measured between two further electrodes (the receiver). The four electrodes are collinear and the distance between the mid-point of the source electrodes and the mid-point of the receiver electrodes is termed the offset. Transient current injection at the source may take the form of a step change in current, such as a reversal in polarity of a DC current, or a coded, finite-length sequence such as a pseudo-random binary sequence (PRBS). For any form of transient current injection, measurements are made of both the source current and the receiver voltage and deconvolution determines the earth s impulse response. Integration of the impulse response yields the earth s step response. Earth Step and Impulse Responses The form of earth response functions may be illustrated by calculating the impulse and step responses at some offset for the simplest case of a uniform, isotropic halfspace. Example impulse responses for land and marine are shown in Figure 1. On land the impulse response comprises a so-called airwave (which travels along the ground/air interface at a scale comparable to the velocity of light and so arrives at time t=0) followed by a response resulting from diffusion through the resistive subsurface. These two components are immediately separable. In the marine case the earth response comprises travel through the sea water, through the sea/air interface and through the subsurface. All three parts persist throughout the entire record. The peak value and arrival time of the peak value (Tpeak) depend on the subsurface resistivity. 17 Geological Society of Malaysia

18 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 14 (a) (b) (c) Figure 1: (a) Land step and (b) land impulse responses calculated at an offset of 1500 m for a uniform halfspace of resistivity 30 Ωm. The earth impulse response has been normalized by its peak value of Ωm -2 s -1. (c) Marine impulse response calculated at an offset of 1500 m for a uniform halfspace of 1 Ωm overlain by 100 m of sea water of resistivity Ωm. The earth impulse response has been normalized by its peak value of Ωm -2 s -1. Note the different timescales. THE EFFECTS OF ANISOTROPY For the transverse anisotropy under consideration (TIV) the vertical resistivity ρ resistivity h define the anisotropy factor ρ v λ = ρ h with typical values between 1 and 5. The square of the geometric mean resistivity is now consider three special ways of varying anisotropy keeping ρ v, ρ h or cases as: c 2 ρ h : ρ h =constant, ρ v and ρ increase with increasing λ c ρ v : 2 ρ c : ρ v =constant, 2 ρ = constant, ρ h and 2 ρ decrease with increasing λ ρ h decreases and ρ v increases with increasing λ ρ v 2 ρ = and the horizontal ρ v ρ. We may 2 ρ constant. We describe these Effects on a uniform halfspace step response for these three cases of varying anisotropy are shown in Figure 2. The effects are dramatic. The airwave (initial step E (0) ) depends only on the horizontal resistivity ρ h (since the airwave is the Transverse Electric (TE) mode) whereas the late time DC value ( E ( ) ) depends only on the geometric mean. Using results from Wilson (1997) for an isotropic halfspace of resistivity ρ 2 ρ ρ h ρ ρ E(0) = and hence =, E( ) = and hence = π r 2π r π r π r This provides a method of determining the anisotropy of the halfspace as h 2 ρ λ = ρ h 1 E( ) = 2 E(0) March 2009

19 DAY 2 SESSION 4 GEOPHYSICS PAPER 14 Figure 2: Effects of anisotropy on step responses at an offset of 2 km for a uniform halfspace. λ =1 (solid black), λ =2 (dash-dotted red), λ =3 (dotted blue) and λ =4 (dashed green). The isotropic case (solid black) is the same in all three cases. Effects on the impulse response for the same models as above are shown in Figure 3. These graphs are the derivatives of those in Figure 2 but the airwave delta function (derivative of the initial step) is not shown. Figure 3: Effects of anisotropy on impulse responses at an offset of 2 km for a uniform halfspace (land case). λ =1 (solid black), 2 (dash-dotted red), 3 (dotted blue) and 4 (dashed green). The isotropic case (solid black) is the same in all three cases. Marine impulse responses are also much affected by anisotropy as shown in Figure 4. The airwave part is seen to depend mainly on ρ h and the earth response (overlapping with the airwave) is more dependent on ρ v. Figure 4: Effects of anisotropy on impulse responses at an offset of 2 km for a uniform halfspace (marine case, water depth = 100 m). λ =1 (solid black), 2 (dash-dotted red), 3 (dotted blue) and 4 (dashed green). The isotropic case (solid black) is the same in all three cases. 19 Geological Society of Malaysia

20 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 14 IMPLICATIONS FOR THE INVERSION OF MTEM DATA We now seek to determine the implications of inverting MTEM data acquired over an anisotropic subsurface with an isotropic inversion routine. Anisotropic data were generated from a model comprising a background geometric mean resistivity of 20 Ωm with an embedded target layer 25 m thick with geometric mean resistivity 500 Ωm whose top was at a depth of 500 m. An anisotropy value λ =2 was used for all layers. Step responses were generated for the three offsets 1.5 km, 2 km and 2.5 km. Isotropic inversions were made for these three offsets individually (single-trace inversion) and for all three simultaneously (multitrace inversion) see Figure 5. Figure 5. Synthetic step responses from the model described in the text (solid black) and isotropic inversions for three offsets singly and in combination. Also shown are responses at each offset from all the inversion models. Figure 5 shows that for any single offset an isotropic model can be found that fits the anisotropic response data. The target in these isotropic inversion models is always shallower than in the original anisotropic model the smaller the offset, the shallower the target. (A survey over a calibration well could be used to determine anisotropy values that yield the correct target depths.) The model derived from inverting the response at one offset was used to calculate the response at other offsets and there are clear misfits (Figure 5). Similarly the use of all three offsets in a multi-trace isotropic inversion failed to produce a model satisfying all the anisotropic data. CONCLUSIONS Inversion at a single offset cannot distinguish between isotropy and anisotropy and can give misleading results concerning target depths. Thus proper interpretation will require anisotropy to be included as part of any inversion scheme. Since single offset data cannot determine anisotropy, simultaneous inversion of multi-offset data will be a necessity. Layer anisotropies will therefore be included as free parameters in forthcoming inversions. Where possible, a survey over a calibration well may be used to determine anisotropy values that yield the correct target depths. REFERENCES Ziolkowski,A.M., Hobbs,B.A. and Wright,D.A., 2007, Multitransient electromagnetic demonstration survey in France: Geophysics 72 (4), F197-F209. Wilson,A.J.S., 1997, The equivalent wavefield concept in multichannel transient electromagnetic surveying: Ph.D Thesis, University of Edinburgh. ACKNOWLEDGEMENTS We thank both Nigel Edwards and the Consortium for Electromagnetic Modelling and Inversion (CEMI) for the use of their forward modeling codes incorporating resistivity anisotropy. March 2009

21 DAY 2 SESSION 6 GEOPHYSICS PAPER 17 OFFSHORE GEOHAZARDS INVESTIGATION - CAN WE DO WITHOUT? Ouzani Bachir 1 and Razali Ahmad 2 1 Orogenic GeoExpro Sdn. Bhd. 2 PETRONAS Carigali Sdn. Bhd. The E&P companies are experiencing challenges with offshore geohazards during their quest for hydrocarbon resources. Offshore geohazards have a direct impact on safety and cost of drilling operations, offshore facilities design installation and production. Offshore geohazards are defined as natural and man-made seabed and sub-seabed features with potential risk to cause damages to assets, environment, health and even loss of lives. Offshore geohazards are grouped into three main categories that are namely: seabed hazards, shallow hazards up to 200m sub-seabed and intermediate hazards up to 1000m sub-seabed. Seabed hazards are investigated using bathymetry, side sonar images, and magnetic. The most common seabed hazards encountered are uneven seabed, hard grounds, fluids escpaes, seabed sediments and man made structures. Among risks associated with seabed hazards are low break up, scouring, pile penetration, land slide, no purchase, buoyancy, inappropriate settlement, free spans, excessive/poor burial, corrosion, excessive tension, conductor hard drive, blocked drilling string, uneven rig emplacement, and low bearing capacity. The seabed hazards and associated risks are presented in Table 1. Shallow hazards are investigated up to 200m sub-seabed using sub-bottom profilers or ultra high resolution seismic. The most common shallow hazards thick soft soils, shallow hard soils, faults, and shallow gas. Among risks associated with shallow hazards are excessive/poor self burial, corrosion, inappropriate settlement, low break, low bearing capacity, differential penetration, punch through, blocked string, conductor hard drive, mud loss, excessive pressure or blow out. The shallow hazards and associated risks are presented in Table 2. Intermediate hazards are assessed using 2D high resolution seismic. These hazards are usually limited to shallow gas accumulations, faults, buried channels and carbonates. The most common risks encountered during drilling are gas blow out, mud loss, and blocked drilling string. The intermediate hazards and associated risks are presented in Table 3. The results of offshore geophysical surveys carried out in the South China Sea revealed that the most prominent geohazards identified are shallow gas, thin hard layers, shallow channels, faults, corals, debris, pockmark clusters, steep seabed, seabed depressions, and mobile sediments. These geohazards are causing every year multi-million dollars losses in assets, delayed projects, rectifications and ultimately loss of lives. But these effects could be reduced considerably by standard geohazards investigation costing only a fraction of probable damages. 21 Geological Society of Malaysia

22 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 17 Table 1: Seabed hazards and associated risks Seabed Hazards Features Hazards Risks Accurate water depth Water Depth Must be less than 80m Boat landing Design Steep seabed gradient Seabed depressions Seabed mound Seabed channels Seabed trench Seabed scarp Uneven Seabed Pipe walking Excessive pipe tension Uneven rig legs emplacement Low break up Obstruct settlement/free spans Scouring around anchors, piles and drilling conductors Scouring and free spans Coral outcrop Rocky outcrop Sonar contacts Boulders Debris Pockmarks Pockmark clusters Gas venting Hard Grounds Fluids escape (Gas/water) Differential rig legs penetration Obstruct rig legs penetration Rig legs run Punch through Obstruct or limited pile penetration Hard drive of drilling conductor and piles No purchase Chain and cable wear Low break up Obstruct pipe settlement/freespans Block drilling string Uneven rig legs settling Excessive rig legs penetration Low break up Pipe excessive self burial Scouring and freespans Sediment buoyancy leading to exposure of piles Low pile bearing capacity Mobile sediments Unstable slope sediments Seabed sediments Scouring around spud cans, drilling conductor, piles and anchors Catastrophic land slide Wellheads Pipelines Platforms Man made structures Major obstruction Obstruct pipe settlement March 2009

23 DAY 2 SESSION 6 GEOPHYSICS PAPER 17 Table 2: Sub-seabed hazards up to 200m and associated risks Sub-seabed Hazards Features Hazards Risks Differential leg penetration Thick very soft CLAY Excessive leg penetration Slope failed sediments Suction forces and hard to extract Low break up Poor purchase Shallow channels Shallow soft soils Excessive anchor burial Low pile/conductor bearing capacity Excessive pipe self burial Partly lithified sediments Probable corrosive effects Pipe wear and freespans Organics Leg sliding Mud loss Very stiff CLAY Cemented carbonates/corals Desiccated crust Dense SAND over soft CLAY Coal Buried boulders Shallow hard soils Differential rig legs penetration Obstruct rig legs penetration Rig legs run Punch through Hard driving of drilling conductor and piles Block drilling string Poor pipe burial and freespans Reactivation Faults Fractures Mud loss Obstruct pipe settlement Biogenic gas Auti-genic gas Shallow gas Differential rig legs penetration Low break up and excessive anchor burial Low pile/ conductor bearing capacity Excessive pipe self burial Excessive rig legs penetration Gas kick Table 3: Intermediate hazards up to 1000m and associated risks Sub-seabed Hazards Features Hazards Risks Buried Coral reefs / Carbonates/limestone Differential leg penetration Excessive leg penetration Deep channels Shallow soft soils Suction forces and hard to extract Low break up Shallow gas Poor purchase Excessive anchor burial Major deep fault Low pile/conductor bearing capacity 23 Geological Society of Malaysia

24 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 18 GENETIC INVERSION: AN INNOVATIVE COMBINATION OF NEURAL NETS AND GENETIC ALGORITHM FOR SEISMIC INVERSION Jumain Marzuki, Jimmy Klinger, Ivan Priezzhev, Trond H. Bo and Gaston Bejarano Schlumberger Information Solutions A new approach to derive an Acoustic Impedance Inversion volume is proposed in Petrel. Multi layer neural networks as well as genetic algorithm are combined together in order to provide a robust and straight forward seismic inversion. Estimation of rock properties using seismic data and derived attributes has always been a very important but challenging task. There are several "schools" using different methods in order to achieve this goal. All of them are based on strong and constraining a-priori information. The required knowledge of an initial model (cf. for the stochastic inversions), or source wavelet (cf. Colored-, Sparse Spike Inversion), is in several cases hard to acquire, if not even impossible. Moreover, the result of this kind of inversion is often biased by the input initial model itself. In the case of Genetic Inversion, the required inputs are limited to the seismic amplitude, and the Acoustic Impedance well logs used as training data. Indeed no single unique wavelet, neither initial property modeling are needed as inputs prior to run the inversion. A genetic algorithm changes the weights of the neural network such that the prediction error is minimized, using principles from evolution. The advantage of this new method to generate property estimation is that the genetic algorithm constrains the convergence of the inversion in a way that the chance to achieve a global minimum error is much greater than in other previous neural network based inversions. Thus, success is quasi absolute. In addition, another huge advantage of this process is that it is not only restricted to conventional Acoustic/Elastic impedance inversion, but that it could be extended to any kind of petro-physical attribute/parameter, which is linked in a meaningful, and straightforward way to the seismic amplitude or derived attribute data. To be more explicit all the parameters contained in the wave-equation are possible candidates such as velocity, density, porosity and bulk modulus. March 2009

25 DAY 2 SESSION 5 COAXING SUBTLETIES FROM SEISMIC BY MEANS OF AN INTELLIGENT INTEGRATED APPROACH A CASE STUDY GEOPHYSICS PAPER 19 Ahmad Bukhari Ibrahim 1, John Ross Gaither 1, Irmawaty Abdullah 1, Vincent W.T.Kong 2, Alexis Carrillat 2, Nina M. Hernandez 2, Md Ramziemran Abdul Rahman 2 1 CPOC, Kuala Lumpur 2 Schlumberger DCS, Kuala Lumpur Figure 1: Amplitude map extraction (left), and corresponding Seismo-Facies (center) and Acoustic Impedance (right) Introduction Much information is contained in the seismic reflectivity signal. The discernment between signal and noise, and the reduction of ambiguity of these signals is a major technical challenge. Interpreting lithology solely on normal seismic reflection strength has been well recognized to have pitfalls. This paper sets a case history of coaxing the subtle relevant data out from the seismic signals by means of an intelligent integrated approach. It is current industry standard, with the availability of fast cost effective computer systems and data storage, to have outputs of a number of seismic data versions besides the traditional full stack seismic data. The ability to cohesively analyze all the available data in a timely manner makes the difference on the economics of petroleum exploitation from exploration through to the development phase. The result of each separate analysis should also converge for the deductions to be considered valid for interpretation. Methodology In this study, the Seismo-Facies Classification was performed to provide qualitative geomorphologic information about potential reservoir distribution, such as channel sand bodies for exploration purposes. This qualitative analysis was done in parallel to the 3D ISIS Global Simultaneous AVO Seismic Inversion to provide an alternative approach to seismic data analysis and focused on extracting some of the textural information contained in the seismic data. The seismic data textural content is usually not directly captured in the inversion process. The value of traditional Hilbert transform attributes (Taner et al. 1979) have been demonstrated in many case histories as well as waveform classification using neural networks, where the seismic trace is decomposed into components of amplitude and frequency information such as in VRS spectral decomposition attributes (Sønneland 1996). 25 Geological Society of Malaysia

26 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 19 There are number of evident constraints when performing classification of seismic attributes, related to inherent limitations of surface based waveform analysis methods. These constraints are summarized as follow: Usually, Seismo-Facies Classification results are limited to the vicinity of existing interpreted or time shifted horizons. Horizon based Seismo-Facies waveform analysis is sensitive to the quality of the seismic pick in terms of signal consistency. The latter makes the results even more sensitive to noise in low signal to noise data conditions. This paper demonstrates a recently developed new classification scheme that addresses these limitations by generating a 3D volume-based unsupervised multi-attribute analysis that captures seismic stratigraphy, structural and/or texture information as a discrete set of Seismo-Facies. In this exploration and appraisal framework, the availability of interpreted horizons would imposes constraints on a traditional horizon-based approach of seismic attributes analysis, which is avoided by going on a full 3D approach. In addition, subtle variations of rock property or depositional features can be overwhelmed by structural imprint and imaging issues on amplitude data. These limitations can be attenuated or completely eliminated when data is reduced to a representation of discrete facies. This Seismo-Facies Classification approach is focused on geometrical attributes also called texture attributes (Randen et al, 2000). These texture attributes describe spatial and temporal relationship of the seismic signal in a small neighborhood. For instance, lateral continuity measured by semblance is a good indicator of a continuity and discontinuity. Because texture attributes capture 3D seismic signature and their relative changes, they can assist in the recognition of depositional patterns and associated lithology. There are essentially two main categories of 3D texture attributes (Carrillat et al., 2002). The first includes texture attributes that portray kinematic features of the seismic traces such as local orientation, signal discontinuity or unconformity. The second category includes generic texture attributes that capture dynamic features in the seismic signal such as spectral representations or amplitude behavior. Here, a combination of both is used to capture the Seismo-Facies variability of the seismic data. Alternatively, the Seismo-Facies texture volume can be selectively combined with suitable standard Hilbert s transformed based attributes in a hierarchical scheme to evaluate the amplitude sensitive response within the textural facies. Both approaches were evaluated in this study. The Seismo-Facies volume realizations were performed using unsupervised seismic attribute classification, using K-mean clustering, (Coléou et al. 2003) had two key objectives. The first was to characterize seismic reflection patterns or seismic textures in the seismic volume, and the second was evaluate the degree or level of detail that could be expected knowing that data quality is moderate in terms of signal to noise ratio. The parameters and signal characteristics of a group of seismic reflectors characterize a Seismo- Facies, which differs from that of a neighboring set of reflectors. The lateral and vertical distribution and associations of the Seismo-Facies reveal patterns that can be interpreted in 3D space and provide information about the geology and/or the depositional environment. In parallel, the Well logs, the interpreted structure surfaces, and the angle seismic partial stacks are processed through the 3D ISIS Global Simultaneous AVO Seismic Inversion route. Careful attention was paid to enhance the quality and lateral uniformity of the input angle seismic partial stacks to generate high fidelity rock property cubes that are calibrated to the wells within the project area. The rock properties data cubes could be used to infill the Seismo-Litho facies geobodies such that variations of pore-fluid or reservoir qualities can be analyzed. This is termed as Hierarchical Seismo-facies analysis. A number of the input well logs are used as constraints to determine the lithology cut-offs, with their associated acoustic rock properties. The determined acoustic rock properties are used as input in a two dimensional Bayesian type simulation to generate the lithology cubes as previously defined. March 2009

27 DAY 2 SESSION 5 GEOPHYSICS PAPER 19 Results The Seismo-Litho Facies approach was driven by a two step procedure, where first an unsupervised classification and evaluation of the results was done against key target horizon and amplitude data. Then a second pass was run, focused on fine tuning the preliminary results to improve features extraction such as channel mapping. The Seismo-Facies Volume 1 was made of a mix of textural attributes (Flatness, Chaos, Frequency and VRS) and amplitude sensitive attributes (Reflection Strength and Gradient Magnitude). This Seismo- Facies volume 1 was then evaluated in attribute space to understand the meaning of the facies in terms of seismic response and was visually inspected along key stratal surfaces and horizons. Next, a filter was designed and used to delineate the main contours of the key channels (Figure 2). The channel delineations were then evaluated on cross plot of different seismic attributes to identify the signature of the channel in attribute space (Figure 3). This analysis was repeated iteratively at different reservoir levels with the short list of attributes in order to assess the best attribute combination based on discrimination power. This operation was used to guide the final selection of relevant attributes for fine-tuning the classification. For example, a channel has a narrow signature in Flatness and Reflection Strength cross plot (Figure 3), which suggest good discrimination power from these two attributes. Conversely, it has a broad signature on the Instantaneous Frequency and Reflection Strength crossplot. This Final Seismo-facies volume confirmed all known channel features observed on amplitude data and provided more shaper definition of them and also provided new indications of a channel belt relevant for additional reservoir intervals. The output data cubes from the 3D ISIS Global Simultaneous AVO Seismic Inversion also were analyzed to compare with the lithological features captured through the Seismo-facies method. The rock properties data cube gave further insights into possible pore-fluid effects and reservoir quality of such features as channels. The Lithology cube showing the simulated occurrence of coals, sands and shales was generated (Figure 4). Conclusions In this study, the seismic input data is piped through two independent routes; the Seismo-Facies Classification using the full offset stack seismic data, and, the 3D ISIS Global Simultaneous AVO Seismic Inversion followed by Lithology prediction. The Seismo-Facies Classification approach is focused on texture attributes that describe spatial and temporal relationship of the signal for qualitative seismic geomorphology analysis. On the other hand, the ISIS Global Simultaneous AVO Seismic Inversion is focused on the elastic properties determination for quantitative reservoir quality prediction. The results from each of these routes are compared, to ascertain the convergence of lithology features coaxed out from their respective route. A few of the hitherto unmapped sand bodies were identified References Carrillat, A., Randen, T., Sønneland, L. & Elvebakk, G Automated mapping of carbonate mounds using 3D seismic texture attributes. Society of Exploration Geophysicists, Expanded Abstracts, 21, Coléou, T., Poupon, M. & Azbel, K Unsupervised seismic facies classification: A review and comparison of techniques and implementation. The Leading Edge, 22, Randen, T., Monsen, E., Signer, C., Abrahamsen, A., Hansen, J. O., Saeter, T. & Schlaf, J Three-dimensional texture attributes for seismic data analysis. Society of Exploration Geophysicists, Expanded Abstracts, 19, Sønneland, L., Tennebø, P., Gehrmann, T. & Yrke, O D model-based Bayesian classification. Society of Exploration Geophysicists Expanded Abstracts, 13, Taner, M.T., Koehler, F., & Sheriff, R.E Complex seismic trace analysis: Geophysics, 44, Geological Society of Malaysia

28 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 19 Figure 2: Example of filter design on the amplitude map for channel body delineation Figure 3: Evaluation of the channel signature in attribute space based on the filter defined on the amplitude map and overlaid on the background attribute response March 2009

29 DAY 2 SESSION 5 GEOPHYSICS PAPER 19 Impedance SHALE SAND COAL Acoustic Impedance Lithology Cube Figure 4: Comparison of Acoustic Impedance transformed into Lithology cube 29 Geological Society of Malaysia

30 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 21 RAPID MULTIPLE-SCENARIO DEPTH STRUCTURE RISK ANALYSIS CASE STUDY IN CUULONG BASIN, VIETNAM. Hong Shien Lee (speaker) and Nguyen Xuan Nam Landmark Graphics, Level 75, Tower 2, PETRONAS Twin Towers, Kuala Lumpur, Malaysia Summary Depth conversion uncertainty is a critical factor for drilling fractured granite basement targets with highly deviated well trajectory in Cuu Long Basin, Vietnam. This paper illustrates depth conversion using 3D velocity model that calibrates many sources including seismic velocities, time depth functions, well picks and surfaces. Various assumptions of velocity interpolation employing different data inputs were evaluated systematically in order to see their effects on the model. Four sensible assumptions were chosen to generate what if scenario velocity models to capture the uncertainty. Case study done on this area using the method described above has produced the base case or most likely case, the high case and the low case. The result suggests that actual basement depth is deeper than previously predicted. The previous velocity model used was created based on single time-depth function from a nearby well. Introduction Cuu Long Basin, a NE-SW trending extensional basin on the southern shelf of Vietnam, was formed during the rifting in Early Oligocene. Late Oligocene to Early Miocene inversion (Hung L.V and Hung N.D., 2003) has intensified the fracturing of granite basement and made it an excellent reservoir. The basement tops of current oil fields are about m deep, in which the oil columns range from m. Basement top of the deepest discovery is at 3700m. Most fractures inside the basement are of high dip angles (40-75 o ). Their strike directions vary from one field to another, or even within a field. Figure 1 shows a typical fractured basement prospect and the key overburden elements: 1- Low velocity in the shallow formations BI and C; typically compaction controlled, and 2- High velocity in E & D formations has variation thickness probably lateral velocity changes. Exploration drilling target is at the productive fractured zone of 100ms time thickness (250m-300m) below Top Basement. The best practice is to design a highly deviated trajectory well perpendicular to the major strike of the fractures, 70 o -80 o inclination inside basement and target about 200m below Top-BSMT. Figure 2 (inset) shows the case study area, exploration target is at basement high prospect X; located Northern part of prospect A that was drilled in the late 1990s, but in the same SW-NE structural trend with prospect B drilled recently. A proposed location of exploration well on prospect X is about 12km away from Well A and 18km from Well B. The previous time-depth conversion was done simply by using only a time depth function from Well B. In order to select optimal drilling targets and design optimal well trajectory as described above, in addition to more accurate depth model, a measure of uncertainty in depth was also required. 3D velocity model was created to calibrate different data inputs: seismic velocities, structure control and time-depth functions from 2 wells A & B. Each input data contributes its characters to the final interpolation velocity model which causes the changes in velocity and consequently the changes in depth structure. This paper describes a methodology for evaluating multiple what if velocity models scenarios to capture the uncertainty/risk of the workflow. For example: what if seismic velocities contain anomalies, what if time-depth function is too fast or too slow, what if there is a bad well tie or what if the velocities are more compaction controlled so that they may not follow structure, etc. The methodology demonstrates how multiple scenarios are systematically developed based on input data and the calibration/interpolation methods applied. March 2009

31 DAY 1 SESSION 2 GEOPHYSICS PAPER 21 Methodology The methodology used was constrained within a short time frame during the final period of drilling program. Applying vertical stretch depth conversion is an accurate method as the overburden structure is not complex and the horizons are relatively horizontal. Data Selection: Velocity model has been generated using the following input data: seismic velocities, time-depth functions at the wells, and time horizons: 1- Seismic velocity used is PSTM migration velocity with 50m x 50m grid. 2- Time-depth functions from 2 offset wells, A & B 3- Four time horizons, Top-Basement and 3 overburden Top-B1, Top-C and Top-D. Another input would be well picks associated with time horizons, which has been excluded from this study due to time constrained and data availability. Model Calibration: Calibration of seismic velocities to time-depth functions uses a calibration volume determined in the interval velocity domain. This is achieved by comparing the velocities at each well with the seismic velocities in the same location, and calculating a calibration function that would correct the seismic velocities to the well velocities. Calibration functions from all wells are then interpolated, potentially along structure, to produce a calibration volume, which then multiplied with the seismic velocities to produce a model that ties the timedepth functions and honours the trends in the seismic velocities. The final calibration is applied to tie surfaces with its associated well picks. What-if scenarios Figure 2 shows the variation of velocity inputs at three wells location: velocity in well A is faster than well B while new exploration well X location has the fastest seismic velocity. How do the observed velocity variations affect the depth structure and its uncertainty? The answer is to build multiple what-if" scenarios velocity model, followed by multiple depth realizations. There are wide range choices of combining input data and velocity interpolation methods to build model. Using Landmark DepthTeam Express velocity modeling package, by simply turning-on or turning-off each of the input data, many what-if scenarios velocity models can be created to capture the different time to depth conversion results. Most importantly, only scenarios that affect velocities changes and depths structure are selected. It is expected that at least four scenarios would be tested for one horizon. The range of typical scenarios includes: Use of time-depth function: manually omitting time-depth curves - a traditional way to analyze the impact of using faster and slower time-depth. Use of Seismic Velocity: Cases were run with and without seismic stacking velocities to consider the impacts seismic velocities on the model. Seismic velocities have less accuracy; although it captures broad velocity trends in the field. In areas of rapid lateral velocity variation the move-out of CMP gathers can be significantly distorted, resulting in false high and low velocities. Use of surfaces (Structural control): this generated a lateral-guided model. The choice to calibrate along structure can make a significant difference. This scenario would apply if the velocities were more compaction controlled, such that they may not follow structure. Use of well picks: picks are hard data, most reliable input data for velocity modeling. Each pick must be associated with a time-surface. Turning-on and turning-off the use of picks helps to identify the reliability of picks and eliminate the bad picks or bad horizons can be done easily. Additional considerable scenarios are: Velocity error interpolation alternatives and using wellderived velocity gradients in the layers instead of seismic velocities Results These scenarios are deterministic. It is possible to generate statistical model volumes from these cases to produce a probabilistic distribution. However, maintaining discrete cases allows for easier update of individual cases as further data or interpretation becomes available. Eight depth structure maps were 31 Geological Society of Malaysia

32 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 21 compared for uncertainty analysis and four were choose to be base case, high case, low case, and fourth case as demonstrated in Figure 3. The High Case is defined as producing the shallowest structures. This case used no seismic velocities but time depth functions from Well B, which has slowest velocity with structural interpolation, and considered that prospect X located at the same structure trend with prospect B. This is very similar to the case using a Well B time-depth function only as Top Basement time structure of two prospects is almost same. The Base Case or most likely velocity model uses time-depth functions from Well B, structural control and seismic velocities. Compared to the case without seismic velocity, the results have significant different. The variation resulted by seismic velocity changes away from the wells is assumed to be caused by geologic variations in the subsurface. It is 100m offset from Base Case to High Case (Figure 3). The Low Case that could be considered as the deepest structure used all available data together: timedepth function from both Well A and Well B with structural control and seismic velocity. This is considered as the standard single output of conventional velocity model. In the crest of the structure, this is similar to the case applied Well A time-depth function with structure control. The differences are only in the flank side as the seismic velocities tended to pull up the structure. Low Case is approximately 100m below Base Case, and 250m to High Case (Figure 3). A Fourth Case was provided using time-depth functions from both Well A and Well B together with structural control. The fourth case falls between the range of Base case and the Low case (Figure 3). The use of deterministic velocity models built using different assumptions that interplay in sometimes unexpected ways does not always lead to a simple distribution of possible depth maps. This underlines the complex nature of velocity modelling. Figure 4 shows the depth map comparison of 4 chosen cases (High, Base, Low, 4th) in which structure shape and closure area varies from one to another. A well was drilled in X structure after this study. The well result confirmed top basement depth is in between the range of 4th Case and Low Case, approximately 40m below 4th Case and 20m above Low Case. Conclusions Velocity model scenarios are effective for capturing uncertainty in depth conversion resulting from multiple data inputs. Each scenario captures different choices made in assembling the input data, where no one choice is necessarily correct. Using tools that allow for rapid model building and updating enables collaborative velocity modeling that benefit from the knowledge of the whole asset team. This methodology can help to improve the quality of interpretation, as it provides a fresh perspective to integrating geological and geophysical interpretations. In this case study, many what-if velocity models were generated to define the high case, base case and the low case of depth structure. The previous prediction using a single time-depth function is equivalent to the High case, which is the shallowest case. Uncertainty was measured at 250m difference from Low to High case. Understanding uncertainty of depth structure helps to finalize well planning and drilling program with much greater confidence. In this study, seismic velocity is key factor to the quality of velocity modeling due to long distance from prospect location to the existing wells. The Low Case velocity model used seismic velocity together with wells time-depth tables and structural control, results in the least error to the actual depth. Variation in seismic velocity observed commonly within D & E formation suggests further research. The latest PSDM studies in Cuu Long basin that focus on analyzing these velocity changes and the geometry of D & E formation were proven to improve imaging of Top Basement structure as wells as the fracture system inside basement. Acknowledgments The authors wish to thank Halliburton/Landmark for their sponsorship and permission to publish this paper. We aslo thank Dylan Mair Landmark Asia Pacific for his continuous and extremely efficiency guidance regarding scenarios-based velocity modeling workflow. March 2009

33 DAY 1 SESSION 2 GEOPHYSICS PAPER 21 Figure 1: A typical a fractured basement high structure and key overburden elements: 1- Low velocity in the shallow formations BI and C; 2- Higher velocity E & D formations probably have some lateral variations. Overburden structure is not complex and relatively horizontal. Figure 2: Map of new basement high (inset) & velocities of each well, prospect X, A and B. The distance from proposed Well X to Well A is about 12km and 18km to Well B. Well X has highest velocity among 3 locations while well B has the slowest velocity. 33 Geological Society of Malaysia

34 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 21 Figure 3: Uncertainty analysis on velocity maps to define High case, Base Case Most likely and Low case from multiples scenarios. Figure 4: Maps compare of multiple-scenario depth maps: structure shape and closure area are various from one to another March 2009

35 DAY 2 SESSION 6 GEOPHYSICS PAPER 23 GEOMECHANICAL MODELLING, SEISMIC PORE PRESSURE PREDICTION AND WELLBORE STABILITY ANALYSIS: KEY ELEMENTS DEMONSTRATED BY AN OFFSHORE EXPLORATION CASE STUDY Adrian White 1, Sunil Nath 1, Katharine Burgdorff 1 and Norbert van de Coevering 2 1 GeoMechanics International, Level 40, Tower 2, PETRONAS Twin Towers, Kuala Lumpur City Centre, Kuala Lumpur, Malaysia. 2 CGGVeritas, Level 56, Tower 2, PETRONAS Twin Towers, Kuala Lumpur City Centre, Kuala Lumpur, Malaysia. A robust, field-specific geomechanical model (in situ stress, pore pressure and rock properties Figure 1) has proven to be an essential tool for the hydrocarbon industry. Constraining the geomechanical model for any particular area provides valuable information for improved characterisation of the region, field or reservoir. The geomechanical model can also help to support a development plan that optimises drilling and production operations. The benefits of incorporating a geomechanical approach into the well planning phase include predicting pore pressure, fracture pressure, wellbore stability and minimising the use of excessive mud weights that may lead to formation damage. The construction of a geomechanical model utilises a broad range of geological, geophysical and engineering data. These data allow the quantification of the magnitudes and orientations of the in situ stresses across the field. The pore pressure can be measured in the reservoir and estimated in shale using wireline log or seismic velocity data. The magnitude of the minimum horizontal stress (Shmin) is inferred by interpreting pumping pressure tests such as leak-off tests. The vertical stress (SV) is calculated from density log data. Figure 1: The geomechanical model is made up of five components. The magnitudes and orientations of the stresses comprise the in situ stress tensor. The remaining components of the geomechanical model are the pore pressure and the effective rock strength. The effective rock strength is composed of several physical rock properties. The magnitude of the maximum horizontal stress (SHmax) is the only stress magnitude that cannot be directly measured. Hence, modelling is required to constrain its magnitude. One of the best ways to constrain the magnitude and orientation of SHmax involves the identification of drilling-induced wellbore failure, such as tensile fractures and wellbore breakouts, through the analysis and interpretation of wellbore image data or caliper logs. A geomechanical model was constructed to limit the risks associated with wellbore instability in an exploration prospect, offshore Malaysia. The aim was to determine drilling constraints, from a geomechanical perspective, for further exploration and development. The construction of the geomechanical model for the prospect can be summarised as such: 35 Geological Society of Malaysia

36 Petroleum Geology Conference and Exhibition rd March, 2009 Kuala Lumpur Convention Center, Kuala Lumpur, Malaysia GEOPHYSICS PAPER 23 A generalised lithology was established using the gamma log and the mud log. Pore pressure profiles in the offset wells were created using wirelinelog based predictive methods. Normal compaction trends were fitted to the sonic, resistivity and density logs and the relationship developed by Eaton (1972) was used to predict pore pressure from the sonic and resistivity data and the Equivalent Depth Method was used to predict pore pressure from density data. Interpreted pore pressures were calibrated to direct measurements, where available. The SV profile was calculated by integrating density and pseudo density from sonic data. The pseudo density profile was calculated using the equation developed by Gardner et al. (1974). The magnitude of Shmin was constrained using reported leak-off tests that did not reach fracture propagation, or tests where time Figure 2: Output from GMI PressCheck showing normal compaction trends fitted to seismic interval velocities, compressional sonic, density, and resistivity data along the path of an offset well. Far right panel shows the predicted pore pressure using the data. The final pore pressure profile was calibrated to RFT data. based pressure plots were not available. The magnitude of Shmin with depth was constrained using the Matthews and Kelly (1967) approach, the Eaton method (1969) and using a minimum horizontal effective stress ratio which assumes the ratio between the minimum horizontal effective stress (Shmin Pp) and the vertical effective stress (SV Pp) remains constant with depth. The absence of image or caliper to interpret wellbore failure meant that the magnitude and azimuth of SHmax were constrained using offset field experience. The depth-continuous rock strength profile was determined using empirical relationships based on sonic data. The equation developed by McNally (1987) was used for sandstone and the equation developed by Horsrud (2001) was used for shale. The potential hydrocarbon bearing sands in the exploration prospect had not been tapped by the offset wells at the time of the study. The shortage of data from these deeper sands meant that the pore pressure had to be constrained by other means. An accurate pore pressure profile for the planned well would directly impact on the accuracy of the wellbore stability analysis. Careful seismic processing is crucial as the standard processed velocities showed a large discrepancy with the sonic data. The mismatch would have followed through into the pore pressure predictions and could have led to erroneous mud weight and casing design recommendations. The carefully processed seismic interval velocities were shown to produce a good match with sonic data from offset wells thus providing confidence in their ability to reliably predict pore pressure (Figure 2). The seismic processing that was undertaken included a residual high-resolution radon de-multiple step, dense bi-spectral velocity analysis on the near offsets only, conversion from RMS to interval velocity using a skeleton which is derived based on the signal-to-noise ratio and structural filtering of the interval velocities along the locally derived seismic dips. These new interval velocities, and the resulting velocity cube, were used to produce a 3-dimensional pore pressure model that can be used to establish pore pressure in any future wells drilled in the field. March 2009

37 Figure 3: Stress summary for the planned well. The summary utilises the stresses established using offset well data and the predicted pore pressure derived from seismic interval velocities extracted from along the planned wellpath. DAY 2 SESSION 6 GEOPHYSICS PAPER 23 The geomechanical model for the exploration prospect indicates that the area is associated with a strike-slip stress regime (Shmin < SV < SHmax). The absence of direct indicators of the orientation of SHmax means inferences on this stress orientation need to be made based on nearby fields in the region. The azimuth of SHmax is assumed to be approximately 081 N. It is important to assess how uncertainties associated with the azimuth of SHmax, due to the lack of image data and caliper logs and the inferences made, impact on the wellbore stability in the planned near-vertical well. Interval velocity traces extracted along the planned well path were used to predict pore pressure. These data show that the pore pressure is expected to ramp up in a step-wise fashion from approximately ~1500 metres depth up to a maximum of ~14 ppg at the total depth of the well (Figure 3). A wellbore stability analysis was conducted for the 12¼ and 8½ hole sections of the planned well with the aim of defining the operating mud windows. These are the range of mud weights bounded by the minimum mud weight required to prevent borehole collapse and maximum mud weight to prevent loss of circulation. For the 12¼ hole section, the mud window is ppg and for the 8½ hole section the mud window is ppg. As a final part of the wellbore stability analysis, quantitative risk assessments (QRAs) were performed to investigate which variables in the prospect specific geomechanical model are most sensitive to uncertainties. The QRAs take the form of Monte Carlo simulations. These assessments show that the magnitudes of the pore pressure and SHmax have the greatest impact on the likelihood of successfully drilled the planned well. The QRA also shows that uncertainties associated with the azimuth of SHmax do not have a strong influence on wellbore stability in the planned well. It may be beneficial to collect additional data during future drilling campaigns to help better constrain some of the more critical parameters in the geomechanical model. A number of recommendations concerning data collection came from study: Acquiring high resolution LWD and/or wireline sonic and resistivity image data could better quantify the azimuth of SHmax if wellbore failure is observed. Modelling the widths of breakouts seen in LWD images can also help to further refine the pore pressure profile. Performing rock strength tests on core samples would help to calibrate the rock strength profile that is currently constrained using wireline data and empirical relationships. Conducting extended leak off tests (XLOTs) or minifrac tests would provide data that could better quantify the magnitude of Shmin. 37 Geological Society of Malaysia