An assessment of the performance of derivative based data enhancement techniques in the presence of coherent noise

Abstract

Enhancement of potential field datasets using operators based on one or more of the spatial derivatives is common practice. The performance of these methods in the presence of noise is poorly understood; other than a general acceptance that they can be significantly affected, especially when higher order derivatives are used. Most published descriptions which involve noise tests use random noise and a dense and uniform sampling of the test region. More realistic tests of the effects of noise should account for the incomplete and anisotropic sampling comprising most datasets and also correlated noise such as due to incorrect levelling. An understanding of the effects of noise on the different methods of enhancement is particularly important when working with lower quality (older) and lower resolution datasets. Interpretation of geophysical data from West Africa, as part of a major project on the prospectivity of the region, is being undertaken. Much of the data available is of relatively low quality and resolution. An important component of the work will involve determining how best to enhance the gravity and magnetic datasets. For the aeromagnetic datasets with wider line spacing, the calculation of dy proved especially challenging with the resulting noise propagating in to enhancement products which rely on this parameter, e.g. tilt derivative. Significantly better results are obtained when dy is calculated from the two other spatial derivatives using Hilbert transform. Our results suggest that line, rather than grid based, processing is most effective for magnetics with grids being created as late as possible during the processing sequence.