A seismic cube can contain a lot of useful information to guide 3D facies modeling and 3D petrophysical modeling. Geophysicists, for decades, have been developing many techniques to correlate signature in the seismic cube to specific facies or log signature along the well. The general goal is to identify a signature in the seismic cube at the well location and to track such signature between the wells, assuming that wells drilled there would see the same log/facies signature than observed on the existing wells. The purpose of the present paper is not to detail all these geophysical techniques – it would be impossible. The reader can refer to books such as (Chopra and Castagna, 2014), (Li, 2014) and (Simm and Bacon, 2014) for some introduction on the topic. Our goal here is to see, once the geophysicist has created one or several new seismic attributes, how shall we use them in our geomodeling workflows?
On a geomodeling point of view, seismic attributes and petrophysics are similar. You can have numerous seismic attributes in the same way that as you can have many different types of logs to model in 3D. But at the end of the day, all these logs are modeled using a limited set of geomodeling techniques and, as explained hereafter, all these seismic attributes are integrated in the geomodel using a limited set of techniques too.

Figure 2 illustrates how a continuous seismic attribute can be transformed into a set of cubes of facies probabilities. Along each well, the facies distribution is compared with the values of the seismic attribute (Figure 2, upper part). In our example, one can see that for values of the attributes up to A, the facies on the well is always sand. We can assume that between the wells, everywhere where the attribute is in that range, we will have always sand. In these cells, the cube of sand probability is equal to 100%, while the cube of shale probability is equal to 0%. In a symmetric way, for values above B, the facies at the well is always shale. In the cells with this range of values, the cube of sand probability is equal to 0% and the cube of shale probability is equal to 100%.

If this binary behavior (either 100% chance to be in sand or 100% chance to be in shale) was true for all values of the seismic attribute, we wouldn’t even need any geostatistical computation. For a given cube of seismic attribute, the facies distribution would be purely deterministic. It would mean that the seismic attribute on its own would be enough to characterize the reservoir. In practice though, the relationship is never as clear as this and there is always at least range of values for which there is probability to be either in the sand or in the shale. This is where geomodeling and geostatistics find its place: it allows throwing the dices and generating multiple facies realizations which respect that information. Concretely, for each attribute value between A and B (Figure 2, upper part) we assign the proper probability in the sand and the shale cubes.

This illustrates what was discussed in the introduction of this chapter. Seismic cubes can’t see the reservoir at the resolution that wells can. As such, seismic attributes can’t capture the details and there are always some discrepancies between the seismic information and the well information. Geostatistics, through the use of cube of facies probability, allows studying this uncertainty. The same approach would work if the seismic attribute was a discrete property.
As mentioned earlier, several attributes might have been computed. Each attribute will generate its set of probability cubes. If multiple sands and shale probability cubes exist, they can be combined into an integrated single cube. The same can be done to merge together a set of cubes coming from seismic and the set of cubes generated from merging VPC and facies proportion maps. The idea is that each cube, each data, shows one aspect of the facies distribution and to understand the whole distribution, we need to combine them.

If a seismic attribute relates to a continuous well log such as porosity, or fracture density, the seismic attribute is directly used as a secondary variable with geostatistical algorithms such as Collocated Sequential Gaussian Simulation (collocated SGS). The workflow is identical to the one described in the paper on petrophysics where the 3D porosity model was used as a guide to model the water saturation in 3D. Such algorithms are able to use multiple secondary variables if needs be.

The reader can refer for example to (Doyen, 2007) for more details about these different techniques.