The spatial and temporal characteristics of observed CO2 plumes obtained from 4D seismic surveys can be honored by finding the most probable models in which the migration of CO2 plumes are spatially and temporally similar to the observation. A computationally efficient scheme is necessary to assess the dissimilarity between CO2 plumes simulated over a large suite of geologic models, and subsequently to select the subset of models exhibiting characteristics similar to the observation. The Euclidean distance is the most common way to measure dissimilarity between vector representations of CO2 plumes. However, the Euclidean distance yields an average measure of dissimilarity between plume shapes and does not take the spatial relation between vector elements into account. For this reason, we measure the shape dissimilarity between CO2 plumes using the Hausdorff distance. In our examples, CO2 plumes selected using the Hausdorff distance are more similar to observed CO2 plumes than those selected using the Euclidean distance. However, it is impractical to conduct complex flow simulations on a large suite of geological models in order to find the most probable models honoring the spatial characteristics of an observed CO2 plume. We save about 80% of the computational cost of selecting the most probable models by screening the large suite of geological models using scaled connectivity analysis. The paper demonstrates the combined use of scale connectivity analysis and assessment of dissimilarity using Hausdorff distance on realistic field examples.
All Science Journal Classification (ASJC) codes
- Industrial and Manufacturing Engineering
- Management, Monitoring, Policy and Law