Publication Date:
2020-02-12
Description:
During the processing of magnetotelluric (MT) data, frequency-dependent, complex transfer functions between magnetic (B) and electric (E) fields are calculated. The transfer functions provide information about the conductivity of the subsurface and thus are of crucial importance. In some cases the data processing proves to be difficult, since the recorded time series can be heavily contaminated by anthropogenic noise signals, e. g. galvanic currents or near field sources. Several methods, like robust or remote reference processing, address these problems, however in case of coherent noise sources they might fail (Junge, 1996). Therefore a new multivariate processing scheme based on an eigenvalue decomposition method (Egbert, 1997) was developed within an AMT study in the Westerwald, Germany (Hering, 2015). The results are presented for frequencies between 10 Hz and 5 kHz. The choice of the noise model is crucial for noise being coherent between different channels at a local site but incoherent to that at remote sites. For an unfavorable signal-to-noise ratio, however, the results of the eigenvalue analysis might be misleading, e. g. if the two dominant eigenvalues are taken as an indication for homogeneous source fields. Furthermore the magnetic and electric field polarizations from the Westerwald data set were analyzed. The results show distinct preferential directions and may be related to artificial source fields. As a consequence the far field assumption has to be checked for subsequent data interpretation.
Language:
English
Type:
info:eu-repo/semantics/conferenceObject
Format:
application/pdf
Format:
application/pdf
Permalink