Parameter estimation for laboratory porous media (PARTILAB)

PARTILAB (PARameter esTImation for LABoratory porous media) addresses a central challenge in porous media research: effective estimation of physical parameters from laboratory flow experiments. This process involves solving inverse problems by tuning model parameters within physical constraints to minimize the misfit between experimental data and model predictions. A valuable side effect of this approach is the quantification of residual modeling error, providing a means to assess and verify modeling capabilities. PARTILAB will focus particularly on high-resolution spatio-temporal tracer data (essentially movies), which are becoming increasingly accessible in experimental porous media research. 

Building on recent advances by the collaborating French and Norwegian teams, a novel parameter estimation framework for laboratory porous media flow will be developed. This framework will be based on the concept of using optimal transport distances, e.g., the Wasserstein distance, to quantify the misfit between data and model. The French team has demonstrated the significant advantages of this approach over traditional methods in seismic imaging. Given the complexity of typical porous media flow patterns like fingers, optimal transport distances offer promising potential to enhance parameter estimation in this domain as well. To enable this, an efficient and accessible computational framework will be created, integrating previous developments from the Norwegian team. The French team's expertise in efficient optimal transport algorithms will be crucial to this effort. 

Furthermore, with access to laboratory data on CO₂ storage and fracture flow, the framework will be tested and validated using real-world datasets. Leveraging the complementary expertise of both teams, this project will both result in scientific advancement of parameter estimation for laboratory porous media research as well as building an important network for future collaboration.