We have many sources of uncertainties in our reservoirs. A wide spectrum was covered in previous papers of this series, through the angle of how we can integrate them in our geomodeling workflow. The present section is meant as a summary of all of them.
The bulk-rock volume is controlled by the geometry of the top horizon, the bottom horizon and of the faults. As such, uncertainty on these different surfaces is to be considered.

If only wells are available (no seismic), one should consider the uncertainty in the contouring far from the wells, and also the uncertainty in the picks themselves. The former is covered in chapter 3. This is the main source of uncertainty. If contouring was done using geostatistical techniques, uncertainty in the variograms and the distributions should be considered too (chapter 2). The uncertainty on the picks is usually minimal, but it might still be good to check, just in case.

If a seismic interpretation is available, very likely no contouring will be applied anymore. Instead, we now have to consider the uncertainties linked to the seismic interpretations: the interpretation itself in the time domain and the effect of the time-depth conversion. The former is a geophysical issue, which goes beyond the scope of this series, while the latter is covered in section 5 of chapter 6.

In faulted reservoirs, the geometry of the faults is a large source of uncertainty. It might even go as far as questioning the presence, or not, of some of them. In such cases, it might be interesting to build some models without the questionable fault(s) and some with.

If several fluid zones are present, the geometry of the fluid contact surfaces is also source of uncertainty. If all the contacts are only apparent (water-up-to, oil-down-to…), the depth of the contact is only partly known. Beyond this, if the reservoir is compartmented, different contacts might co-exist.

The 3D property models are also a large source of uncertainty. The one most often ignored while it is, in fact, a key factor is the internal geometry of the 3D-grid itself. Building the internal mesh, for example, parallel to the top horizon, or parallel to the base, or parallel to another surface might create some very different property models. It all relates back to uncertainty on the deposition space within each geological unit (chapter 1).

Beyond that, uncertainty in property modeling lies within the choices we make in term of geostatistics (chapter 2). What algorithms are we using? What values for the input parameters? Facies proportions are a key controller of the oil volumes in the reservoir. Facies proportions are also an input parameter to many algorithms. It is important to build models based on variations of these proportions. Uncertainties in the variogram shape, size and orientations will mostly lead to uncertainty in the level of connectivity of the reservoir rocks. For the oil sands, for example, this is important as we tend to consider only the volumes of large connected geobodies (in which we can apply steam-assisted production techniques).

Facies tend to have specific petrophysical ranges of values and these ranges tend to be narrow within each facies. As such, uncertainty in the porosity or the So values tend to be less of an issue. If you have to choose, focus first on characterizing the uncertainty on the facies. But if time allows, look also at the possible uncertainty in the logs themselves (chapter 4).

All in all, uncertainties in bulk-rock volumes and in facies (proportions, orientation, but also the geological hypotheses about the depositional environment) tend to be the main ones impacting the range of volumes. However, only a review, with the whole team, will allow confirming it or pinpointing to other more important ones in your specific reservoir.