Many critical natural resources - such as fresh drinking groundwater, fossil fuels, geothermal heat, or mineral resources that are required to produce sustainable energy - are exploited underground. This exploitation is often costly, difficult, and subject to high uncertainty because the exact position of these resources is not accurately known in 3D. To minimize the risks, it is crucial to develop probabilistic models allowing decision makers to evaluate the risks, to decide about additional data acquisition, and to avoid wasting natural resources inefficiently.
Stochastic simulation approaches for resources evaluation can be classified into two main families, depending on whether they are model-driven or data-driven. The former relies on the specification of a parent spatial random field, characterized by the set of all its finite-dimensional distributions. For example, truncated Gaussian simulation defines the geological domains by truncating a Gaussian random field and can be applied when these domains obey an ordered sequence that can be confirmed through the geological setting. Plurigaussian simulation is an extension of truncated Gaussian simulation that allows working with more complex contact relationships between geological domains; in particular, the modeling process can explicitly specify the allowed and forbidden contacts and the chronology between domains. In contrast, data-driven approaches do not rely on a genuine spatial random field model, but on statistics and spatial continuity measures that can be inferred from the borehole data or from a training image (TI). This derivation is deemed representative of the spatial arrangement of the geological domains in the subsurface. For instance, multiple-point statistics algorithm enables the reproduction of more complex geometries and spatial patterns.
In this context, we propose to launch a long-term collaboration between the Center for Hydrogeology and Geothermics of the University of Neuchâtel, Switzerland and the School of Mining and Geosciences in Nazarbayev University, Kazakhstan.
The aim of this collaboration will be to develop joint research activities in the broad field of uncertainty quantification for natural resources. The research will include theoretical and mathematical aspects related to uncertainty quantification, the development and exchange of software, their test and applications to various problems in Kazakhstan and Switzerland, and the development of guidelines for best practices. To start, we will focus on those two common data and model-driven simulation algorithms, namely, Plurigaussian simulation (PLURISIM) and multiple-point statistics (MPS). Two early researchers (a Swiss and a Kazakh MSc or PhD students) will be involved. They will compare the MPS and PLURISIM uncertainty estimation techniques for two typical situations inspired from real applications: one will be related to properties of geothermal reservoirs in Switzerland, the other will be related to oil reservoir issues in Kazkhstan.
More generally, the project will set up a network of scientists from both universities to exchange experiences related to uncertainty quantifications for natural resources. Meetings and tutorials will be organized to share practical experience. The two young scientists will spend several months of exchange in the two partner universities. Finally, time is allocated for the preparation of a joint research proposal to allow the development of a long term and in-depth collaboration.
Prof. Philippe Renard, University of Neuchatel
Dr. Nasser Madani, Nazarbayev University