Exploiting Full Impedance Tensor Data with Three-dimensional Magnetotelluric Inversion

Wednesday, 17 December 2014
Nate Lindsey1, Edward Alan Bertrand2, T. Grant Caldwell2, Erika Gasperikova1 and Gregory A Newman1, (1)Lawrence Berkeley National Laboratory, Berkeley, CA, United States, (2)GNS Science, Lower Hutt, New Zealand
Experience modeling magnetotelluric (MT) data in a complex three-dimensional (3D) environment shows that analysis of the full impedance tensor is necessary, since using only the principal component (off-diagonal) data can produce artifacts in the imaging process. While analysis of only the off-diagonal components might be justified when the imaging grid is aligned with preferred structural trends in the data, for truly 3D environments this may be difficult if not impossible to achieve. Nevertheless, full MT tensor analysis has its own challenges, especially how one treats data noise in the inversion. For example, some suggest weighting the data with fixed percentages of impedance; however tensor elements can vary by several orders of magnitude at a single period. In these instances, the accuracy of error estimates in the data could be used as they contain useful information regarding measurement uncertainty. We investigate these technical aspects of data noise and appropriate weighting using a large MT survey (259 soundings over 1250 km2) recorded across New Zealand’s Taupo Volcanic Zone. These data have been acquired to explore the deep roots of high temperature geothermal systems, which are thought to arise from convective plumes that extend down to the brittle-ductile transition (~6-8 km depth). However, there is still much to learn about basement structures, permeability and the mechanisms of heat transport below depths of 3 km, which is the present maximum drilled depth. Treatment of full impedance data in the MT inversion process is used to help address these questions in our investigation.