Inversion of SkyTEM data

SCI constraints setup
1D-model types

Constrained inversion

Inversion of the airborne EM-data in Aarhus Workbech is preformed with the AarhusInvcode.  Locally we use a 1D model description/EM-response.  The models are laterally and spatial constrained forming a pseudo 3D model spaces. The inversion code has be customized to handle vary big airborne EM surveys and support multi CPU cores with a very high parallel efficiency.

Besides the resistivity models provides the inversion code also a depth of investigation value (DOI) and a data residual  for each resistivity model.

  • LCI-setup: The models are laterally constrained along the flight lines forming at 2D model space.
  • SCI-setup: The resistivity models are constraints laterally along the flight lines and across the flight lines, resulting in 3D constrained model space. A full SkyTEM survey can be inverted as one single SCI setup.

The laterally constraints can either work in depths or elevation.

Model types

The smoothness or sharpness  in the model results is control by the strength of the constraints and the regularization scheme. Aarhus Workbench offers three main type of model discretization/ regularization scheme:

  • Smooth: The resistivity model is discretize using several layers (~30) with fixed layer boundaries. The regularization penalizes vertical changes in resistivity, resulting in a vertical smooth resistivity model.
  • Sharp: The resistivity model is discretize using several layers (~30) with fixed layer boundaries. The regularization penalizes the number of vertical resistivity transitions of a certain size, resulting in resistivity models with relative sharp vertical resistivity transitions. The sharp regularization scheme can also be applied for the laterally constraints.
  • Layer: The resistivity model is characterised by few number of layer (~4-5). Both layer thickness and resistivity are model parameters. No vertical regularization is applied which result in distinctly layered resistivity models with a fixed number of layers true out the survey. A built in routine estimates the numbers of layers needed to fit the dataset based on a smooth model inversion results.

Prior constraints

Aarhus Workbench also supports prion constraints on any model parameter. The prion constraints can be initialized from grids and direct from the GIS map or be specified at borehole locations with decreasing  strength moving away from the borehole locations.

Accurate system modelling

To obtain high quality inversion results is it important to modelling the airborne EM-system in great details. We therefore include the following parameters in the modelling:

  • Transmitter, receiver heights
  • Shape of transmitter loop
  • Transmitter waveform
  • System low pass filters
  • Front gate filter
  • Width of the individual time gates