Department of Geoscience

Workpackage 2

WP2. Voxel inversion of geophysical data for improved hydrological integration.

WP leader: Professor Esben Auken (AU)


Geological and groundwater 3-D models most often refer to a regular grid, often referred to as a voxel-grid. Inversion of electromagnetic geophysical data has migrated from single-site inversions to profile-oriented inversion using lateral constraints (Auken et al., 2008) to inversions including entire surveys using spatial constraints to obtain geologically reasonable results (Viezzoli et al., 2008). These all refer to a model space being linked to the actual observation points. This means that incorporating the geophysical data into the regular geological and/or hydrological modeling grid involves a spatial relocation of the models, which in itself is a very subtle process where valuable information can easily be lost. For frequency-domain data Brodie and Sambridge (2006) developed a grid-inversion method in which they calibrated and inverted the data in a regular grid. Lately a three-dimensional grid inversion for AEM data emerged (Cox et al., 2010), but one-dimensional codes are still warranted for large surveys.

Within the HyGEM project there have been developed a geophysical inversion algorithm working directly in a regular voxel grid disconnected from the actual measuring points. The voxel grid comply with the model space of the hydrological model, eliminating loss of information. As an important sideeffect, the algorithm can then handle dramatically larger data volumes with inclusion of a priori information from boreholes. The code will be a further development of in-house software (AarhusInv program package (Christiansen and Auken, 2008)) at Aarhus University being used for large AEM surveys (Christiansen and Auken, 2008; Viezzoli et al., 2008).