Knowledge to application: meta data approaches to improved geological model conditioning in petroleum industry workflows

Sedimentary rocks are commonly highly variable internally. For example when rivers leave behind sediments, channel deposits may sit within floodplain deposits, such that ribbons of sand become encased within a muddy background. To describe and study sedimentary systems such as these, sedimentologists use descriptive terms such as "channel" or "floodplain" to recognise building blocks that together stack up to build a rock volume. These building blocks are called architectural elements. Geologists also recognise that architectural elements may themselves stack up in organised patterns, and such organised patterns of elements can be placed in hierarchical arrangements. For example, channels may stack together to form channel complexes, and channel complexes may build channel complex sets. This hierarchical descriptive approach can work in the deposits from a wide range of sedimentary environments: e.g., rivers (fluvial rocks), shallow seas (shallow marine rocks) and deep seas (deep marine clastic rocks).
Sedimentary systems built from pre-existing particles (such as sand grains, or mud grains, that might make sand- or mud-stones) are known as clastic systems. In the right circumstances, such systems may form hydrocarbon reservoirs. This happens when the sediment is buried, but with connected pores between the grains still open (i.e., without minerals growing in the pores during burial to seal the rock). If oil or gas migrates into connected pore spaces, but cannot migrate out, because the rock above is sealed in some way, the hydrocarbon filled rock volume may have the potential to be an oil or gas reservoir, if there is enough hydrocarbon in place, and the predicted flow rates are high enough.
To predict whether a known hydrocarbon accumulation might make an economic reservoir - i.e., be worth developing - oil companies model its behaviour as a possible field. They may first build a sedimentary model, recognising different architectural elements and the way they are stacked together. They will then account for any deformation the rock experienced during or post deposition. Finally they will simplify this geological model to build a reservoir simulation model, in which the performance of the possible field can be predicted. Ideally the company would build the geological model using seismic reflection data that showed the basic geology (this is a remote sensing technique that builds 2- or 3D images of the subsurface geology based on processing the reflections of sound waves sent into the ground). However, the resolution of these techniques is usually not good enough to show exactly what the subsurface geology is like, and the companies have to use models to fill in the data they can't see directly.
To build these models, companies commonly use computer techniques to generate synthetic geology at the finer scale, using algorithms that randomly generate patterns of architectural elements based upon modelling rules. However, problems often arise because these rules are not always based upon the way the geology actually tends to stack together in particular settings. The database approach is a new way of determining what these organisational rules should be. They provide data that are more reliable, because the data are all compiled from real world examples of geology. We have already applied this approach to rocks deposited by rivers, and in the deep sea, and many companies have used the results in their own modelling. However, the approach hasn't been tried yet for shallow marine rocks - and that's what we aim to do in this project. Shallow marine rocks host many oil and gas fields, so if we can improve the modelling of such fields, we'll have a significant impact upon the efficiency of the companies who use the technique, as we'll reduce the uncertainly they commonly experience when deciding whether or not to develop a field, and how to extend the lives of fields that are already producing.