Vector and total field data inversions for magnetometer UXO surveys in the North Sea

Abstract

As an alternative to the industry standard processing of total field data, a Basin-hopping inversion can be used. Another alternative is the use of vector field magnetometers, which can be processed with a Gauss-Newton inversion. Both the Basin-hopping and Guass-Newton inversion consist of a two step inversion. First, an inversion to estimate the centre location and magnetic dipole moment of a point source is carried out, the so called point-source inversion. Second, the retrieved magnetic dipole moment can be used to reconstruct the magnetic moment with a prolate spheroid model, which is called the prolate-spheroid inversion.

First, both the TMI and vector data inversion are tested on 3000 noise-free synthetic datasets located in the North Sea where a prolate spheroid with varying shape, orientation and magnetic susceptibility creates an anomaly field. The largest global root mean square (GNRMS) error for the point-source inversion of the Gauss-Newton vector data inversion has a value in the order 10−9 whereas the Basin-hopping inversion for the TMI data returns for only 61% of the cases a GNRMS error in the order 10−7 or lower.
The result of the point-source inversion serves as a plateau for the prolate spheroid inversion; when the point-source inversion does not reach the set level of accuracy, it cannot be determined if the prolate spheroid model describes the true object well. For both the TMI and vector data, the UXO was correctly classified in 72% of the 3000 datasets. 
Second, comparing the current industry standard to the Basin-hopping point-source inversion only gave inconclusive results. The estimated centre location of both methods are within a 0.15 m radius but the disagreement in magnetic moment leaves an advice on which technique would work better indecipherable.