# 7.4 Shall we model the net-to-gross?

To some extent, in the general volumetric equation, the net-to-gross is mathematical trick used to ignore the portion of the bulk rock volume which is not in the pay zone; either because it is in the water zone or because it is made of non-reservoir facies. While porosity and So are properties we can observe or measure along the well, the net-to-gross can’t. It is really an intellectual construct.

With this in mind, shall we model the net-to-gross in our geomodeling workflows with geostatistical techniques, the same way we do it for facies, porosity or So? Mathematically-speaking, we could, and sometimes indeed we should. But a lot of thought must be applied to understand what this NTG 3D property is going to represent in a given reservoir.

Net-to-gross can be expressed as a binary log along each well. For all the measured depths belonging to the pay zone, the net-to-gross log takes the values 1, otherwise 0. Computing the net-to-gross for the whole well, as defined in the traditional volumetric workflow, can then be done simply by counting the number of “1” values along the log, multiplied by the MD resolution of the net-to-gross log.

All the geostatistical techniques used to model facies can be used to model any discrete log, like the net-to-gross log. One could apply a geomodeling workflow in which the petrophysical properties would be distributed by net-to-gross values (“0” and “1”) instead of doing by facies.

Is this approach recommended? Modeling net-to-gross can be necessary, on occasion, as explained in the last paragraph of this section, but with a slightly different geomodeling workflow. In most cases though, it is safer to avoid doing this.

The key issue is that the net-to-gross variable is based on several reservoir characteristics which can have independent trends. Once “hidden” within the net-to-gross quantity, it becomes impossible to model these trends properly. The problem might occur even for a simple reservoir.

Let’s imagine a thick reservoir with no water zone (sealed faults are delimiting the reservoir) and no cut off needed on porosity nor So. Net-to-gross is only defined based on facies: sand is reservoir; shale is not.

Let’s imagine that, in fact, we have two facies sand, one for sand deposited in channels and one for sand deposited in sand bars. These two sands will have different spatial distribution because they are from different depositional systems. If we model facies, we have three facies, one shale and two types of sand, and we can apply geostatistical algorithms and parameters specific to each sand. But instead of modeling facies if we are modeling the net-to-gross, then we are losing the information about two depositional systems. It would be impossible to distribute the pay net-to-gross values (“1”) while respecting the geological constraints in the same way we could for the facies.

Things get even more complicated for reservoirs in which the net-to-gross is also based on porosity cut-off. The net-to-gross is now hiding possible different trends for each facies as well as specific porosity trends potentially even specific to each facies (due to vertical compaction or effects like sand grain coarsening or thinning). Add on top of this new net-to-gross cut-offs based on So or even on pay continuity (ex: ignore pay locally if it’s less than X meters thick) and modeling net-to-gross with geostatistical techniques while respecting the real characteristics of the reservoir can get really, really tricky.
For that reason, it is wiser to take the time to model the facies and then the petrophysics by facies. Then, once the 3D model built, one can always apply a formula to compute locally a net-to-gross property from the different petrophysical properties. As with the wells, the net-to-gross is now derived from the rock characteristics; it doesn’t guide how they should be distributed.

As mentioned earlier though, in some cases, a sort of net-to-gross property might be needed too. It can be used to capture heterogeneity which is of smaller resolution than the vertical cell size of the geomodel. In such approach, the net-to-gross won’t take only values “0” and “1”, but any number between 0 and 1.
For example, let’s imagine a thinly laminated sand reservoir in which the succession sand/shale is at the centimeter scale. Ideally, we would need to model such a reservoir with a 3D grid of vertical sub-centimeter resolution. Mathematically, it would be impossible though: the 3D grid would have hundreds of millions of cells and computations would take forever. We are obliged instead to use a coarser resolution (let say 10cm). If we were to simply to create a facies in the 3D grid with values “sand” and “shale”, we know that we would make a mistake: in fact, each so-called “sand” cell would have a certain percentage of shale in it and each so-called “shale” cell would also have a certain percentage of sand.

A continuous net-to-gross property could help capturing this though and the geomodeling could go as follow:

• No facies modeling
• Compute a continuous net-to-gross property in the cells crossed by the wells. A value of 1 would mean the cell is 100% made of sand while a value of 0 would mean it has no sand in it. Every ratio in-between is possible. Model this continuous net-to-gross property with geostatitics.
• Independently from this, model in 3D the porosity and the So using only the log values from the thin sands as input.
• Compute the volumes per cell and then sum them over the whole 3D grid (same workflow than before).