Separation filtering of aeromagnetic data using filter-banks

Abstract

Filter-bank transforms analyse the local frequency content of a signal. This information can be used to design matched filters that adapt to local properties of the data. Matched filters can be applied to aeromagnetic data to separate the responses of depth-varying magnetic layers and are known in this case as separation filters. The performance of filter-bank separation filters is demonstrated on two synthetic data sets, each with a deep and shallow magnetic layer with differing amounts of depth variation. The modelled data were analysed with a Gaussian filter-bank and conventional global Fourier methods. The Gaussian filter-bank method was more effective than the Fourier method when the depth to the layers was highly variable because the filter-bank was able to adapt to the changing local frequency properties of the data. The numerical accuracy of the Fourier method helped it perform better when the layers were at approximately constant depths because the observed data closely fitted the Fourier model. These results give a preliminary indication that the filter-bank approach to separation filtering provides an effective method for separating the anomalies of magnetic layers that vary significantly in depth. The method also automatically produces crude depth estimates of the layers that may be useful as a first-pass interpretation.