A high-resolution spectral study of audiomagnetotelluric data and noise interactions

This study is an investigation of the behaviour of audiomagnetotelluric (AMT) response estimates, in the presence of intense anthropogenic noise. The details of the data and noise interactions are studied by constructing highly resolved AMT response functions using the Maximum Entropy spectral analysis method, in its multichannel form. For low-noise, consistent data spectral estimates and their frequency resolution should be independent of the spectral technique used. Such behaviour is confirmed. When the data are contaminated by significant amounts of noise the highly-resolved estimates display detailed inconsistencies as a function of frequency. Such piecewise distortions can be understood if the response function is viewed as a polynomial. The results obtained enable the identification of response structure generated by both harmonic and subharmonic noise processes. High-resolution estimates provide a detailed view of data quality and narrow-band perturbations. The comparison of high-resolution response functions with conventional (discrete) estimates, shows that, in some cases, the latter may produce erroneous estimates. A discussion of the generation and received characteristics of noise waveforms is also included. A smoothing operation, involving frequency domain convolution, is described and applied to the results obtained. The technique exploits the minimum-delay properties exhibited by a valid impulse response function in the time domain.