Variability of geologic materials in combination with the scarcity of data lead to a statistical or geostatistical description of subsurface systems coupled with stochastic simulation. Uncertainties in model formulation and model choice, in boundary conditions and external system forcing as well as uncertainties of material parameters have to be considered. Even assumptions that characterize unknown spatial variability of subsurface materials must be considered as uncertain.
Without appropriate consideration of these uncertainties, overly confident predictions, engineering designs and management decisions are inevitable, followed by legal and financial liability.
Experimental site data, e.g., from well tests or tracer tests, may be assimilated into stochastic simulation via Bayesian updating. This helps to reduce the prediction uncertainty. A frequently asked question is which kind of site investigation will lead to the most informative data or most confident engineering design at the smallest investigation costs.
These stochastic modeling tasks are notoriously expensive in computer power. Task-driven problem formulations, suitable simplifications and problem-adapted analytical or numerical solution techniques need to be found and combined in order to make the stochastic applicable to complex real-world engineering tasks.
Typical fields of application include, e.g.,
- the stochastic definition of well capture zones in heterogeneous or karstic systems,
- the prediction of human health risk from groundwater contamination in the vicinity of drinking water wells,
- cost minimization of site exploration and remediation actions through optimized tradeoff between safety margins and additional data acquisition
- the design of CO2 injection schemes with a focus on system failure.
Focal point of research activities
|Laufzeit: 16.06.2011 bis 15.06.2014
|Ein probabilistisches Risikomanagement als integrales Trinkwassersicherheitskonzept