A Remote-reference-free Parametric Method for Robust Impedance Tensor Estimation from Limited Time Period Data

Abstract:

In practical applications involving impedance tensor estimation, robust estimation with remote reference processing (RR) is the prevailing approach to mitigate estimation bias and other issues caused by noise contamination found in real magnetotelluric (MT) data. In contrast to conventional spectral-analysis-based, non-parametric approaches, the proposed parametric approach aims at determining the small number of autoregressive-moving-average (ARMA) model parameters optimized to best agree with a given MT data time series. Our prior studies have focused on validating the parametric approach in mathematical studies and controlled numerical experiments based on simulated 1-D MT data. In this study, we investigate the application of the parametric approach to measured MT data from USArray sites MNF34 and VAP58. The results present apparent improvement in bias reduction of transfer impedance estimates and the best fit accuracy of 81% for the validating geoelectric field prediction, which demonstrates that the impedance tensor estimated with the parametric approach and a data window of one day is similar to the RR estimate obtained using a data window of one month. Given the demonstrated statistical efficiency of the parametric approach, it is possible in certain situations to estimate the MT impedance tensor from site data with far fewer data points than would be required for robust estimation using a conventional, non-parametric approach.