The SEM - revisited (April 1998)

Ian Bryant of Schlumberger Doll Researchdescribed how the new reality of the SEM is being developed.

The basic problem is that the reservoir is unevenly and under sampled. As an example, the area actually sampled by logs in an oilfield may be as little as 0.0001%. The impact of this depends on reservoir geometry. For flat lying beds you may get away with relatively low sample density. For a labyrinth type reservoir this is unlikely to hold true. Current 3D models honor some of the data but introduce a new problem. That of an implicit confusion between real data and interpretation. There is a requirement to visualize what we know and where we know it; and what and where we don't know. Typically costly processing and interpretation may be performed on some datasets such as well logs or 3D seismics. But neither high resolution log information, nor 3D stratigraphic information actually gets into the model. At a well, complex reservoir information may be collapsed to binary sand/shale voxels - while seismic information may be reduced to top and bottom of reservoir. "If a picture is worth a thousand words, an image is worth as many wiggles".

So is the SEM Visualization software revisited? No according to Bryant who describes the use of the SEM in the "validation gauntlet" whereby a fast simulator is used to predict and match iteratively to obtain a number of models that fit the data with an accompanying measurement of associated uncertainty. This was demonstrated with a video of a 3D view of a fault-block in Statoil's Gullfaks field. In one window a ray-traced seismic model was compared with the recorded seismic data. In another the input geological model could be tweaked and the impact of such adjustments viewed in real time on the seismic simulator. The demonstration was sufficient to impress some Statoil personnel in the audience who may have had some second thoughts of the deal they have just struck with Landmark for an enterprise-wide computing solution.