Model-based separation filtering of magnetic data

Abstract

Separating the fields produced by sources at different depths is a common requirement in the interpretation of potential field data. Approaches to this problem are generally data- or model-based. Data-based methods require clear linear segments in the logarithmic power spectrum of the data corresponding to different components of the field. Various types of filters can then be designed to carry out the separation. When the logarithmic power spectrum shows no identifiable linear spectral segments, other approaches are necessary. We outline a model-based method that does not depend on power-spectral information but requires independent estimates of the average depths of the source distributions, e.g., from seismic interpretations. An ensemble of models using fractal source distributions is computed based on these known values, and filter parameters are determined that produce the closest fit (in a least-squares sense) to the theoretical fields that each source distribution generates. This approach is used to separate basement effects from intrasedimentary sources in magnetic data collected over the Colville Hills, Northwest Territories, Canada. Seismic data interpretation places crystalline basement at ∼10 km depth and an intrasedimentary basaltic layer at ∼2 km. Our approach results in an optimal separation filter with a cutoff wavelength of ∼12 km that appears to provide an effective separation of the two source effects.