Fast Three-Dimensional Interpretation of Magnetotelluric Data.

In this paper we present the results of the comparative study of two techniques for 3-D magnetotelluric (MT) data interpretation: the iterative Born inversion and the quasi-analytical inversion. The Born iterative method is based on the fact that the EM inverse problem is actually a bi-linear problem, a special case of a general nonlinear problem (Chew, 1990; Portniaguine and Zhdanov, 1999). Since the problem is linear to anomalous currents, and the currents are the product of the total electric field by anomalous conductivity, the inverse problem becomes linear with respect to conductivity if the total fields are fixed. The total electric fields, in turn, are connected to anomalous conductivity via the corresponding integral equation (IE). Using these properties, one can construct an iterative process whereby the linear inverse problem is solved first for anomalous currents, assuming the total electric fields are fixed. Second, the values of the total fields are updated using the found values of anomalous currents. These new values of the total fields are used in the next iteration of the inverse problem, etc. Another method uses the quasi-analytical approximation for forward modeling (Zhdanov and Hursan, 2000; Zhdanov, Dmitriev, Fang, and Hursan, 2000). This approximation provides fast but accurate technique for electromagnetic forward modeling operator and for the Frechet derivative calculations, which simplifies dramatically the forward modeling and inversion. Iterative Born method and quasi-analytical inversion technique are implemented using the focusing inversion, which provides both smooth and focused images of the geoelectrical structures. The comparative numerical study demonstrates that both 3-D methods work extremely fast and produce reasonable images of the subsurface geological formations. They give us a possibility to invert on a PC within less than half an hour massive magnetotelluric survey data using fine discretization grids, with several tens of thousands of cells. The comparative study of two 3-D inversion methods were conducted using model data and the observed MT data collected by INCO Exploration.