5.4 Seismic interpretation and structural modeling
The technique presented in the previous section, to tie depth-converted seismic horizons to well markers, can be applied to adjust depth-converted seismic fault surfaces to fault markers. The remainder of this section provides an opportunity to discuss how geophysicists and geomodelers can collaborate to model a fault network through integration of seismic interpretation and geomodeling techniques.
In faulted reservoirs, only seismic data can really help defining the geometry of the faults and define how all these faults are interconnected to form the fault network. A faulted geomodel is only as good as the seismic interpretation of the fault network, and how well this interpretation has been respected. In many projects, the workflow is a two-step process (Figure 1, “traditional workflow”). The geophysicist spends time creating and fine-tuning his interpretation of the fault network. Then, once the interpretation is finalized, it is transferred to the geomodeling package and the geomodeler creates the model. Building a faulted 3D-grid has been a tedious task in the geomodeling industry for years, partly because many packages were using an approach called pillar gridding. This technique is described later in this section as it is still used today. For complex fault networks, such geomodeling tasks can take a lot of time. In that context, a complex fault network would be one with many faults, maybe some faults dying in the reservoir or/and some complex fault relationships such as Y-faults, X-faults. Because it takes time to build such model, it is often impossible to circle back with the geophysicist, even if building the geomodel highlights some inconsistency in the geophysical interpretation. There might simply be no more time left in the project to do these adjustments.
Over the last few years, some geomodeling packages have added more robust and simpler techniques to build quickly a fault 3D-grid. For the geomodelers having access to such modern workflows, it allows them to work with their geophysicist in a more integrated way (Figure 1, “integrated workflow”). The idea is for the geophysicist to provide a good, but not complete initial interpretation. The detailed cleaning, traditionally done to make life easier for the geomodeler, is not needed anymore. The geomodeler builds an initial structural model from this dataset, using the modern structural modeling workflows now available. The goal is to quickly get a 3D-grid which can be reviewed by the whole team. As for the traditional workflow, the review often leads to the need for further refinement of the geophysical interpretation and/or of the geomodel itself. But here, as the initial geophysical interpretation and the initial structural model were built fast, there is still time in the project for the geophysicist and the geomodeler to refine together the structural model. The final 3D-grid represents the reservoir more accurately. For this reason, we highly recommend using such modern approaches whenever available.
Considering that pillar gridding is still used a lot, more details are given hereafter (Figure 2). On the contrary, at the time this chapter is written (autumn 2015) not all the details about these modern workflows have been made public yet by the different software companies. As such, a general idea of what can be achieved with these techniques is illustrated in this section (Figure 3).
Figure 2A shows a faulted reservoir in which the faults F1 and F2 form a Y-fault network. The faults and the horizons A and B have been picked on seismic. The reservoir is also crossed by a well on which facies interpretation has been done (two facies zones are represented in orange and yellow). The 3D-grid represented on Figure 2A is typical of what a pillar gridding algorithm can generates. To the left of fault F1, the vertical mesh is parallel to the fault F1 near F1 and it progressively gets vertical when we reach the left limit of the modeled area. Similarly, the 3D-grid to the right of the fault F2, the mesh is parallel to the fault F2 and it gets progressively vertical the further away we are from it. Modeling the block between the faults is the challenging part. Most likely, and not represented here, the model would have been simplified to make it happen. The fault F2 will have been edited so as to remove the branching. Its geometry might have been altered to move the branching deeper, below the modeled year. Another option could be to make F2 vertical and completely crossing the reservoir (fault dying in the reservoir are also an issue for the pillar gridding algorithm). In some project, the fault F2 might have been simply ignored and the throw it creates on the horizon A would be smoothed out. In all cases, very likely, the geometry of this fault network would have been altered.
Figure 2B explains in further detail why this is happening. The first step in pillar gridding is to connect each extremity of the top horizon to an extremity of the bottom horizon. By extremities, we mean points labelled 1, 3, 9, 10, 5 and 7 on Horizon A and the points 2, 4, 6 and 8 on Horizon B. Point 1 from Horizon A is connected to point 2 of Horizon B because they correspond to the left extremity of the input horizons. Point 3 from Horizon A and Point 4 from Horizon B are connected because they both represent the point of contact between the horizon and the fault F1 on the left fault block. Using a similar approach, the points 5 and 6 are connected as well as the point 7 and 8. But what shall we do with the extremities 9 and 10 on Horizon A? As the block between the two faults doesn’t extend to Horizon B, there is no extremity on Horizon B to associate them to. And that’s where the pillar gridding algorithm is blocked. Simplifying the geometry of Fault F2 is the trick used to make sure we have as many extremities on Horizon A that we have on Horizon B. Once it’s done and all the extremities are connected, the vertical mesh is further defined inside each block so as to make a smooth transition between one connection to the next one. At last, the horizontal geometry to the mesh is added.
The new structural modeling workflows developed over the last few years seem to focus on two things.
Firstly, they automate a lot of the tedious tasks usually done manually in older workflow. These tasks are not described in any details here, except to say that ensuring that the model is sealed has always been one of these issues. By sealed, we mean that there must be no gap between the horizon surfaces and the fault surfaces where the former gets in contact on the later. The gap visible on Figure 2 and Figure 3 between the horizon and the fault surfaces were added only to give more clarity to the pictures.
Secondly, these new workflows try to ensure that there no need to simplify the fault network anymore. It means that they don’t rely on pillar gridding algorithms. Figure 3 is an example of the type of output 3D-grid we can get. The fault block between F1 and F2 is now properly managed. Observe also how the cells of the 3D-grid are now cut by the two faults instead of being parallel to them (Figure 2). This is also an improvement as conceptually, one has to imagine that rocks got deposited before they got faulted (except around growth faults). As such, it makes sense that our facies distribution gets “cut” as in Figure 3 instead of getting aligned to the fault as in Figure 2.
As for the integration of local seismic events, described at the end of the previous section, these modern structural modeling techniques have both challenges and benefits. Challenges, because they are still new and use geomodels and we are not yet used to them. But beneficial too, as some complex reservoirs, when applied correctly, might be the key to building a really good geomodel.