Multisite, multifrequency tensor decomposition of magnetotelluric data

Accurate interpretation of magnetotelluric data requires an understanding o f the directionality and dimensionality inherent in the data, and valid imp lementation of an appropriate method for removing the effects of shallow sm all-scale galvanic scatterers on the data to yield responses representative of regional-scale structures. The galvanic distortion analysis approach ad vocated by Groom and Bailey has become the most adopted method, rightly so given that the approach decomposes the magnetotelluric impedance tensor int o determinable and indeterminable parts, and tests statistically the validi ty of the galvanic distortion assumption. As proposed by Groom and Bailey, one must determine the appropriate frequency-independent telluric distortio n parameters and geoelectric strike by fitting the seven-parameter model on a frequency-by-frequency and site-by-site basis independently Although thi s approach has the attraction that one gains a more intimate understanding of the data set, it is rather time-consuming and requires repetitive applic ation. We propose an extension to Groom-Bailey decomposition in which a glo bal minimum is sought to determine the most appropriate strike direction an d telluric distortion parameters for a range of frequencies and a set of si tes. Also, we show how an analytically-derived approximate Hessian oft he o bjective function can reduce the required computing time. We illustrate app lication of the analysis to two synthetic data sets and to real data. Final ly we show how the analysis can be extended to cover the case of frequency- dependent distortion caused by the magnetic effects of the galvanic charges .