Automatic modelling and inversion for dykes from magnetic tensor gradient profiles - recent progress

Abstract

We have developed an optimization method for automatic dyke delineation from observed magnetic and gravity gradient traverse data. A non-linear least squares algorithm is used to find model dyke parameters that best fit the computed gradient tensor data to the observed data. The eigen-system of the observed magnetic gradient tensor data is used to provide starting model dyke parameters for an iterative non-linear least squares solver. This greatly enhances the ability of the solver to find a plausible dyke model for matching observed and synthetic tensor gradients locally. The method works well on synthetic examples. Multiple surveys using a Full Tensor Magnetic Gradient (FTMG) signal instrument from IPHT, have been made in Southern Africa. A real case study with remanence, taken from the Platreef near Pretoria, shows that the gross observed gradient features can be recovered by our procedure, but the residuals in the gradient fit hint strongly at the need for more complex dyke models. There is more directly inferable structural geology in this tensor signal than can be found in a conventional TMI signal.