3D electromagnetic inversion for environmental site characterization

摘要

A 3-D non-linear electromagnetic inversion scheme has been developed to produce images of subsurface conductivity structure from electromagnetic geophysical data. The solution is obtained by successive linearized model updates where full forward modeling is employed at each iteration to compute model sensitivities and predicted data. Regularization is applied to the problem to provide stability. Because the inverse part of the problem requires the solution of 10's to 100's of thousands of unknowns, and because each inverse iteration requires many forward models to be computed, the code has been implemented on massively parallel computer platforms. The use of the inversion code to image environmental sites is demonstrated on a data set collected with the Apex Parametrics (open quote)MaxMin I-8S(close quote) over a section of stacked barrels and metal filled boxes at the Idaho National Laboratory's (open quote)Cold Test Pit(close quote). The MaxMin is a loop-loop frequency domain system which operates from 440 Hz up to 56 kHz using various coil separations; for this survey coil separations of 15, 30 and 60 feet were employed. The out-of phase data are shown to be of very good quality while the in-phase are rather noisy due to slight mispositioning errors, which cause improper cancellation of the primary free space field in the receiver. Weighting the data appropriately by the estimated noise and applying the inversion scheme is demonstrated to better define the structure of the pit. In addition, comparisons are given for single coil separations and multiple separations to show the benefits of using multiple offset data.