Multiple-station data processing

Magnetotellurics (MT) relies on natural electromagnetic field variations to investigate the electrical conductivity structure of the subsurface. Modern MT data are often multivariate due to simultaneous\nrecordings of multiple-channel time series of two (horizontal) electric and three magnetic field components at\nmultiple stations. Single site and remote reference processing only use a small portion of data to estimate the impedance tensor. The multiple-station approach, initially presented by Egbert (1997), uses all data\ninformation to improve the signal-to-noise ratios, which results in better estimations of the transfer functions.\nThis is particularly import in industrialized regions, where the influence of man-made noise signals often exceeds the natural EM fields and hampers the estimation of MT impedance tensors. We have included the multiple-station data approach in our processing scheme EMERALD and tested it with different data sets. A\nnon-robust calculation of the impedance tensor based on the multiple-station approach shows already slightly improved results compared to robust single site or even remote reference estimators. However, in case of high level man-made noise advances of the multiple-station algorithm are not observed. Tests with existing\nrobust routines within EM, which are based on bivariate assumptions, do not reveal a significant improvement.