Upper drawing: During a typical CSEM survey, a dipole source is towed above EM receivers placed on the seafloor. The presence of hydrocarbon-filled sediments in the subsurface will scatter the EM field, and part of the scattered field propagates back to the seafloor where the signal is recorded by receivers equipped with electric and magnetic sensors. Lower image: To find out if oil or gas is present in the subsurface, acquired EM data must be processed and interpreted. This is an extensive and iterative process that requires access to advanced processing and analysis tools. (Images courtesy of Rocksource)

The concept of CSEM surveys is based on the knowledge that the propagation of an electromagnetic (EM) field induced in a conductive subsurface is mainly affected by spatial distribution of resistivity. In marine environments, saltwater-filled sediments typically represent good conductors, whereas hydrocarbon-filled sediments, salt, volcanic rocks and carbonates represent examples of resistive bodies that scatter the EM field. Part of the EM field scattered by subsurface inhomogeneities propagates back to the seafloor where the signal is recorded by receivers equipped with electric and magnetic sensors (Figure 1, upper drawing).

In the early stages of handling marine CSEM data for hydrocarbon detection, simplified approaches were used to process and display the data (such as the well-known amplitude versus offset and normalized magnitude plots). Such processing certainly still has its merit as it in a simple way reveals trends in the data along the receiver line. However, such an approach suffers from a number of limitations, including single frequency, single offset and single receiver processing. As a result, most companies handling CSEM data have turned to inversion and migration to provide resistivity images of the subsurface that are interpreted together with other geophysical and geological information (Figure 1, lower image). This approach also provides better control on the depth of observed anomalies.

The current challenge associated with CSEM technology is to correctly handle the acquired data. The electromagnetic (EM) response recorded by the receivers must be decomposed into its components, and the appropriate energy levels must be attributed to each contributing factor. This process is well known in the seismic industry. However, the higher level of complexity that governs the propagation of the electromagnetic field through the subsurface forces us to think more creatively. One way around the problem is an integrated approach where geological a priori information is incorporated into an iterative interpretation process. This approach, which has been applied to the current case study, allows fast processing of data up to the level of full 3-D inversion and migration. Through a structured and iterative interpretation process, appropriate values for each individual contributing factor can be obtained. As a result, the complexity of the problem is reduced.

The Luva challenge

In 2003, a CSEM survey was acquired to test the EM response from the Luva gas discovery, which was made by BP in the Norwegian Sea in 1997. This survey was one of many 2-D datasets that the Norwegian company EMGS acquired in the early phase of developing the CSEM method. The purpose of acquiring the data was to test the technology on proven discoveries in a multitude of geological settings.

While most of these lines proved that the technology worked, there was no significant EM anomaly detected over the Luva reservoir. This was quite disappointing because it implied that the CSEM technology appeared to be unreliable.

The authors re-examined the Luva data to try to understand the EM response. An investigation of the acquired CSEM data determined that the quality of the Luva dataset was good. As part of the integrated analysis approach, the reservoir section was studied in detail. Although the reservoir is considered to be of generally high quality, the resistivity log shows significant resistivity anisotropy. The calculated effective resistivity that contributes to an EM anomaly for the reservoir interval is less than 20 ohm, while the surrounding rocks have resistivities varying between one and seven ohm.

Synthetic EM modeling based on a realistic target and background geology demonstrates that the response contrast from the Luva discovery, when compared to background resistivity, will be less than 10%. It becomes apparent that the Luva target is hard to detect mainly due to the low effective resistivity of the discovery, which results in low resistivity contrasts between the reservoir and the overlying shales. In order to solve the Luva case, the ability to handle low-resistivity contrasts in the processing is crucial. This required an advanced integrated approach.

Solving the challenge

The starting point for further analyses involved building a representative 3-D model based on seismic data and a sound geological understanding of the area. Seismic data were particularly important at this stage as they enabled resistivity values to be attributed to particular intervals.

A synthetic 3-D CSEM dataset was then generated and inverted. As a starting point for the inversion, no constraints were used. The results clearly indicate that the Luva hydrocarbon column cannot be expected to be detected using this basic approach (Figure 2, upper section).

The next step was to do a similar inversion with real data as input. The inversion result is very similar to that of the synthetic data, indicating that the geological model is representative of the real geology and further demonstrating that the Luva discovery can not be identified using this approach (Figure 2, lower section).

The next step in the analysis was to provide regional constraints (but absolutely no target constraints) based on a sound geological understanding of the area. Seismic data were particularly important at this stage as they enabled resistivity values to be attributed to particular intervals. We deliberately chose not to use the well data from Luva to create the constraints because prior to discovery such data would not have been available. In order to most accurately replicate an exploration scenario, the well data should be ignored.

By using regional constraints to the modeled data, the inversion was guided towards what the geoscientist considers to be a more appropriate solution. Applying this integrated approach to the synthetic model data indicated that, providing the geological model used was representative of the real geology, the Luva discovery could be identified using this workflow (Figure 3, upper section).

Next, the same constrained approach that was applied to the synthetic model was used, but this time with real data. Again, the inputs to the constraints were taken from regional interpretation of seismic data and a sound geological understanding of the area. No well information was used, and no constraints were used for the target itself (including parameters such as geometry, depth, thickness, lateral extent and resistivity). The final result shows that the Luva discovery is clearly identified with a pronounced resistivity anomaly (Figure 3, lower section). The similarities to the synthetic example suggest that the synthetic model is realistic and consistent with observations from real data.

Conclusions

The integrated approach used to solve the Luva case represents what explorationists are trained to do. CSEM data represent one piece of information, to be used as a part of the puzzle, just as the well and seismic data are. No explorationist would treat seismic data in isolation, and similarly, CSEM data should not be treated independently either.

Furthermore, if the oil and gas industry is to exploit the full potential of CSEM technology, it must understand the complexities involved both in relation to detection and delineation. Basic analyses only form a starting point for the investigation. Advanced processing and inversion are normally required to fully understand the data. In complex settings, it is essential to use all available information such as input from seismic and borehole data as well as our geological understanding that may be based on working a province for years. Properly used, analyses of the CSEM data will improve our geological understanding of the investigated area.

Acknowledgements

We thank GeoExpro (www.geoexpro.com) for allowing us to use the content. The SBL data used in this study was provided to us by EMGS (www.emgs.com). Public domain (DISKOS) seismic data were collected in 1996 by CGG for BP and are maintained by Statoil. The well information was obtained from www.npd.no. The advanced workflow used to solve the Luva enigma is a result of efforts by numerous Rocksource scientists and colleagues to whom we are grateful.

EDITOR’S NOTE: This article is based in part on a case study presented in GeoExpro, 4, 2007, pp. 52-58, and HGS Bulletin, December, 2007, pp. 23-43.