• Experimental design based methods
• Bayesian probabilistic formalism
In general, experimental design methods are based on generating simpler response surfaces using selected rigorous simulations (e.g., of reservoir models with appropriate fully coupled phenomenological process models) followed by Monte Carlo simulations to identify the parameters that are most critical in affecting the targeted outcome. Sufficient data to validate models of trapping mechanisms and associated failure modes will not necessarily be available from any previous as well as ongoing field tests. For example, very few new wells that could provide calibration data, including log and seismic data, will be drilled in most ongoing or planned field demonstrations. In these situations where information is sparse, the Bayesian framework can be used to obtain probability distributions of critical parameters. In this approach several conceptual geologic models are constructed based on all available data. Each scenario is then depicted using a spatial variability model. It is recognized at this stage that several choices of spatial variability models (variograms, for example) are available. The global estimate of a given target variable is constructed using the interpretation of the quantitative data under a given geologic scenario. Once the uncertainty in a given parameter is quantified, experimental design tools can be used to design monitoring programs with an aim of reducing uncertainty. A general framework for determining and quantifying uncertainty in carbon dioxide sequestration applications is built and some examples are shown.