Building and history-matching a three-dimensional model is a difficult procedure. The third dimension increases parameter requirements, model run times, and model output uncertainty. Ideally, predictive uncertainty can be reduced through assimilating information from site characterisation on the one hand, and measurements of system behaviour on the other hand. But this is so much harder in three dimensions than it is in two dimensions.
This tutorial explores some new concepts which have the potential to make three-dimensional, decision-support modelling more fruitful than it is now. It continues the “conceptual points” theme from a previous tutorial. A hydrogeologist draws a “conceptual mudmap” of a study site. This allows a modeller to construct and parameterise a numerical model, to build prior probability distributions for model parameters, and to generate and adjust three-dimensional parameter fields that emerge from these probability distributions. These parameter fields can be efficiently adjusted using PESTPP-IES or PEST_HP.
The tutorial is long, and somewhat experimental, as it breaks new ground. Methodologies are innovative and can be improved; however they are ready for real-world use.
The tutorial covers many issues. Not only does it suggest some new and exciting ways of working in three dimensions. It also compares different history-matching and uncertainty analysis methods. It shows how decision-support modelling can benefit from using them in complementary ways, after understanding the strengths and weaknesses of each.
Topics that are covered by this tutorial include the following:
- Encapsulating the outcomes of site characterisation using conceptual points;
- Generation of 3D, stochastic, non-stationary parameter fields that can paint semi-realistic pictures of subsurface hydraulic property heterogeneity;
- Numerically cheap adjustment of these fields using PESTPP-IES and PEST_HP;
- Data space inversion as a stand-alone means of uncertainty analysis or as a complement to other methodologies.