[Truncated abstract] Long-term predictions of subsurface flow are important for societal issues such as groundwater flow, renewable and non-renewable energy resources, nuclear waste disposal and CO2 sequestration. In complex realistic settings, numerical simulations are dependent on the distribution of below-ground properties defined by geological models, constructed from observed data. The quality of the data therefore directly influences the predictions of subsurface flow. To date, no framework exists that allows a direct evaluation of effects of data quality on these flow fields. This thesis presents a first comprehensive method to analyze, visualize, quantify, and couple geological data uncertainty to flow field predictions. Methods are introduced to simulate and evaluate uncertainties in complex 3-D structural geological models. Based on probability distributions assigned to the underlying data, realizations of the geological model are created with an automated implicit modeling technique for complex 3-D spatial settings. The concept of information entropy is applied to visualize and analyze uncertainties in the resulting geological models. Information entropy values are a measure of the minimum number of geological units that can exist at any point in the domain. In addition, measures of mean model entropy can be used to derive quality estimates of the discretization required for geological modeling. Techniques are described that significantly simplify the integration of complex geological modeling into flow simulation, allowing an automatic update of flow and temperature outputs when data in the geological model are added or changed.
|Qualification||Doctor of Philosophy|
|Publication status||Unpublished - 2011|