This title appears in the Scientific Report :
2013
Towards Large Scale Multi-Layer-Conductivity Inversion of Quantitative Electromagnetic Induction Data
Towards Large Scale Multi-Layer-Conductivity Inversion of Quantitative Electromagnetic Induction Data
Electromagnetic induction (EMI) systems enable high spatial resolution measurements within short times. Multi-offset EMI devices sense different depths and allow in principle a better vertical characterization of the subsurface, but lack in quantitative measurements due to static shifts that occur...
Saved in:
Personal Name(s): | von Hebel, Christian |
---|---|
Mester, Achim / Huisman, Johan Alexander / Bikowski, Jutta / Rudolph, Sebastian / Vereecken, Harry / van der Kruk, Jan | |
Contributing Institute: |
Agrosphäre; IBG-3 Zentralinstitut für Elektronik; ZEA-2 |
Imprint: |
2013
|
Conference: | 73. Jahrestagung der Deutschen Geophysikalischen Gesellschaft, Leipzig (Germany), 2013-03-04 - 2013-03-08 |
Document Type: |
Conference Presentation |
Research Program: |
Modelling and Monitoring Terrestrial Systems: Methods and Technologies |
Publikationsportal JuSER |
Electromagnetic induction (EMI) systems enable high spatial resolution measurements within short times. Multi-offset EMI devices sense different depths and allow in principle a better vertical characterization of the subsurface, but lack in quantitative measurements due to static shifts that occur due to the influence of cables and/or operator. To calibrate the recorded apparent electrical conductivities (ECa) a linear regression between predicted ECa, obtained from a Maxwell-based exact forward model using inverted electrical resistivity tomography (ERT) data as input, and measured ECa is performed.
Recently, a two-layer inversion was introduced, using a combined one dimensional global-local search (GLS). The global-search optimizes along a regular grid using an approximate model. The subsequent local-search uses a Simplex minimization and an exact forward model. This approach uses no smoothing or damping to assure sharp layer boundaries. Here, we extended the GLS to three-layers. Thus the parameters increased from three to five enlarging the solution space and increasing the difficulty to find the global minimum. The GLS was implemented without and with lateral constraint which compared the current optimizations with the parameters obtained prior to that position. Large deviations called a new global and local search before inverting the next position. Moreover, a shuffled-complex-evolution (SCE) optimization was implemented that inverts each position separately using the exact forward model.
Experimental EMI and ERT transect data were acquired at the Scheyern research farm of Helmholtz-Zentrum-München. Performance and reliability of GLS and SCE were tested by running the optimization from start-to-end and from end-to-start of the profile. The GLS inversion results without lateral constraint showed a strong direction dependency indicating that the solution space consisted of too many local minima that trapped the inversion. The constraint stabilized the inversion, but the results still remained direction dependent. The SCE inversion results were direction independent indicating that the global minimum was found. Smoothly changing layer properties were obtained without large lateral jumps. Comparison with ERT inversion results showed similar lateral and vertical conductivity changes. The three-layer multi-configuration EMI inversion based on the SCE optimization is a powerful and widely applicable tool to image subsurface conductivity variations. |