An increasing number of oil companies are recognizing the strength of controlled-source electromagnetic (CSEM) data for mapping fluid content. CSEM acquisition produces data which are sensitive to resistive regions in the earth. The overall conductivity of a porous rock is largely controlled by its fluid content. Since brines are very conductive whereas hydrocarbons are not, the CSEM method offers a good chance of distinguishing between brine-filled and hydrocarbon-bearing rocks.

Figure 1. 1a: Acoustic impedance is plotted for a range of porosities (horizontal axis) and gas saturations (vertical axis). The contours of the AI are more or less vertical in the region of good reservoir (upper right of the plot), showing that AI in this well is mostly sensitive to porosity and not at all to gas saturation. 1b: Elastic impedance for the same range of porosity and saturation. The surface is rather flat, showing that EI carries little information about either porosity or saturation. 1c: Resistivity for the same models. The contours here are almost horizontal, showing that resistivity is sensitive to saturation but not porosity. (All figures courtesy of Rock Solid Images)

Of course, this is not the full story. Massive limestones, salts and volcanics may also show low conductivities. Even in porous sandstones, the presence of clay minerals in the pore space may change the overall conductivity quite significantly. The conclusion is that rock conductivity is influenced by porosity, fluid type and to some extent mineralogy.

There are various rock physics models predicting conductivity from these properties. For example, Archie’s equation relates conductivity to porosity, brine saturation and brine conductivity for clean sands. Modifications such as the Waxman-Smits formula account for the presence of clay minerals.

Similar remarks can be made for seismic data. They have been used successfully for many years to map structure. Amplitude variation with offset (AVO) and acoustic and elastic impedance inversion (AI and EI) are used to obtain seismic rock properties such as elastic moduli and possibly density. These in turn may be related to mineralogy, porosity and fluid properties through numerous rock physics relationships, for example those summarized in “The Rock Physics Handbook” by Gary Mavko and his colleagues. While AVO or AI/EI offer the possibility of fluid prediction, the potential ambiguity and risk of misleading results such as confusing fluid changes with lithology changes is generally higher than that from the CSEM data.

The seismic and CSEM data are controlled by different physics and are sensitive to slightly different aspects of the rock properties. We would like to combine them in a manner which exploits their respective strengths while ameliorating their weaknesses. The key to successful combination is using the rock physics consistently.

Rock physics modeling allows us to investigate the information carried by the different surface measurements. Figure 1 shows an example from the Nuggets prospect in the North Sea. We have used standard methods to calculate seismic rock properties, in this case acoustic impedance and elastic impedance (at 30° incidence angle) as they vary with gas saturation and porosity. For this reservoir, we see that the AI responds mainly to porosity and only weakly to gas saturation. In the areas of good reservoir (higher gas saturation and moderate to good porosity) the EI is represented by a very flat surface, indicating that it contains little information about either reservoir property. On the other hand, the resistivity is controlled almost entirely by the gas saturation. In this particular case, then, we can combine the AI and CSEM data to obtain both porosity and saturation estimates and omit the EI since it is not adding much useful information. Of course, this conclusion is reservoir-dependent, and in other cases we may find different combinations more useful.

As mentioned above, Archie’s equation is often used to predict resistivity from reservoir properties. It is important to bear in mind that it only holds for clean sands. The electrical properties of clay minerals are notoriously complicated, and their presence in pore space may change the effective rock conductivity significantly.

Figure 2. Figure 2a shows the resistivity measured in the well against the value predicted using Archie’s equation. The color coding shows the shale fraction. There is a large scatter of points, and most are located in the lower left of the point cloud, suggesting that the prediction is not very accurate. This is not surprising since Archie’s equation is intended for use in clean sands (the blue points). In Figure 2b, the prediction uses the Waxman-Smits model. This accounts for the presence of clay minerals in the pore space, thus improving the quality of the prediction.
Figure 2 illustrates this. We first compare the resistivity predicted from Archie’s equation with the logged resistivity values. It is clear that the prediction is rather poor, and there is a wide scatter of points and no linear relationship between actual and predicted values, even for the clean sands. The same comparison using the Waxman-Smits model, which accounts for clay minerals in the pore space, shows a more linear trend, albeit with some scatter and some anomalous regions in the crossplot. Overall, though, accounting for the clay content improves the prediction significantly.

Ultimately the predicted electrical and elastic properties are used to generate synthetic data for comparison with the recorded CSEM and seismic data. For this purpose it is essential to build a good-quality model from the target zone all the way back to the earth’s surface. CSEM propagation is essentially a diffusive process, and thus the entire region above the target influences the recorded response. In the seismic case, the key point is to use the correct velocity trend in the overburden in order to model the relationship between offset and incidence angle accurately; otherwise AVO modeling may be misleading.

Figure 3. Acoustic impedance and resistivity logs, and the seismic and inverted EM data on a line close to the well. The CSEM anomalies are the high resistivity red regions in the low-resistivity blue background. The well location is in the center of the seismic/CSEM display. The gas sand has intermediate AI in the well between lo- impedance shales above and high-impedance brine sands below. (Seismic data courtesy of TGS-Nopec)
In the Nuggets study, the main interest is to estimate gas saturation in the reservoir. In the well, the gas sand stands out very clearly as a high-resistivity, low-density layer about 80 ft (24 m) thick (Figure 3). This is seismically resolvable, and we have already seen from the modeling that the combination of resistivity and AI offers a good possibility of obtaining the desired information. A constrained inversion is applied to the CSEM data to produce an image of resistivity with depth, just as the seismic data are inverted to AI. For both types of data, we must perform a well tie to ensure correct depth calibration. Well tie for CSEM data involves finding consistency between the resistivity logs and the inverted CSEM data. Just as with seismic data, the different vertical resolution of the two measurements must be reconciled by suitable scaling processes.

Despite these complications, a complete set of well logs coupled with careful rock physics analysis contains enough information to achieve our aim, at least in a semi-quantitative sense. Calibration at the well allows us to combine the surface data and produce a gas

Figure 4. Semi-quantitative estimate of gas saturation superimposed on the seismic wiggle traces. Blue colors represent high gas saturation, and reds are low saturation. The uncolored regions are outside the bounds considered anomalous for resistivity or impedance.

saturation attribute (Figure 4). This is semi-quantitative: blue values represent high gas saturations, and reds represent low values. The scatter of red points away from the reservoir formation is an indication of the uncertainty in the calibration process. It is clear that the CSEM data have good lateral resolution, marking the edges of the gas-bearing zones quite precisely. Vertical resolution is controlled largely by the seismic resolution.

The Nuggets study demonstrates that careful rock physics analysis permits combination of surface seismic and CSEM data to provide useful reservoir information. The final result could not be obtained with either type of surface data alone. This is a promising start to the process of CSEM, seismic and well data integration, and OHM and RSI are working to further refine and broaden the applications of combined CSEM, surface seismic and well log data.