Less is more – uncertainty in rock properties

My last article opened a topic of uncertainty in basin models. Dis­cussion following its publication brought up an interesting research paper describing how in­creasing model complexity increases uncertainty in its results. It illustrates the issue with several examples span­ning various fields of science…

“Our” field offers another exam­ple. Basin models involve the calcula­tion of rock properties used to calcu­late heat and fluids transfer through a basin. These rock properties mutually control each other through a web of physical relationships. Some of these relationships are expressed in the model.

For example, porosity is a function of sediment mechanical compaction, cementation and dissolution. These processes are controlled by mineral composition, grain shape, effective stress, fluid chemistry and tempera­ture. Each of these parameters needs to be calibrated with detailed petro­graphic data. Such calibration is only possible for selected siliciclastic reser­voirs with drilled wells sampling key lithologies. Then, we are able to mod­el porosities within +/- 1 % of rock volume.

Such detailed calibration is not available for the vast majority of a ba­sin fill. In fact, we do not even know the distribution of the basic lithology types for most of the basin. Thus, po­rosity is typically only calculated with an exponential function of vertical ef­fective stress adjusted for general sed­iment type (sand, shale, limestone, etc.). The uncertainty in such calcu­lated porosity is around +/- 5 – 15 % for a given lithology and a range of depth. This porosity is then used to calculate permeability using some function adjusted for general sedi­ment type.

However, permeability is also controlled by multiple processes and parameters and so the calculated per­meability is associated with an uncer­tainty of
2 – 3 orders of magnitude. Now, combine it with the porosity uncertainty and the permeability un­certainty easily becomes 3 – 4 orders of magnitude. But it doesn’t end here. Finite element models go on and cal­culate capillary entry pressure (Pc) based on the permeability.

I remember a conference pres­entation on such a calculation, dis­cussing minute details of Pc evolution through modeled geologic history. In reality, given the irreducible un­certainties, the magnitude of Pc re­mained largely unconstrained. The calculated Pc was no more accurate than a hand-drawn line on a graph basically saying that Pc tends to in­crease with sediment compaction.

So after all the time, IT and fi­nancial resources spent computing unconstrained details for millions of cells, we arrive at a model with some layers being flow barriers / seals and some being flow carriers… which is about as accurate as a hand-drawn sketch on a piece of paper. The com­plicated model creates an illusion of accuracy and barely allows time for the generation of a couple of scenar­ios.

I prefer a simpler map-based ap­proach, manually assigning prop­erties to horizons based on geologic insight, openly admitting the un­certainty and using the resources to explore various scenarios that can be tested with data and lead to more accurate understanding of questions that matter.