Inverse groundwater modelling and stochastic generation of geologic structures

For a groundwater model to be reliable a good conceptualisation is essential. One of the most important aspects of conceptualisation is the choice of geometry of the geological units to be represented by the model. Generally, use of the traditional methods of zonation and/or geostatistical approaches does not allow for a good representation of highly heterogeneous and complex geologic structures such as alluvial environments and fracture systems. As a proper representation of the geology is fundamental for the credibility of model-produced predictions, we intend to develop an improved method for stochastic generation of more realistic geometries of geologic structures. The method should, among other things, put emphasis on interconnectedness of high conductivity zones which, if not dealt with properly, can cause especially transport-related predictions to become useless. We hope to be able to couple the simulation of realistic fields with conditioning to heads and end up with an improved method for estimation of uncertainty. The study is still in an initial phase.
Ph.D. study by Christen Knudby carried out at Polytechnical University of Catalonia, Spain, and funded by the Danish Research Academy.