A discussion of bias in magnetotelluric responses

Abstract

Magnetic-field noise biases magnitudes of magnetotelluric response functions such as resistivity, impedance, and the induction vector (tipper). This is particularly severe in the so-called dead band between about 1- and 10-s periods. The standard approach to remove these biases is to use a magnetic remote reference. In cases where this is unavailable or does not work because noise is correlated at the local and remote sites, it is possible to extract unbiased estimates using an inversion algorithm that predicts the magnitude from the phase. However, an important condition for this to be successful is that phase estimates are unbiased. Phase can be biased if parallel components of the electric and magnetic fields are correlated. This arises when the structure is 3D or when a 2D response is estimated in coordinates not parallel to strike. Improvements in electric field measurements make electric field references a viable option for estimating low-bias responses up to periods approaching 1000 s.