Cluster analysis of Euler deconvolution solutions: New filtering techniques and geologic strike determination

Abstract

Euler deconvolution often presents the problem of filtering coherent solutions from uncorrelated ones. We have applied clustering and kernel density distribution techniques to a Euler-generated data set. First a kernel density distribution algorithm filters uncorrelated Euler solutions from those consistently located near an anomalous magnetic-gravimetric source. Then a fuzzy [Formula: see text]-means clustering algorithm is applied to the filtered data set. The computation of cluster centers reduces the size of the data set considerably, yet maintains its statistical consistency. Finally, the computation of eigenvectors and eigenvalues on the cluster centers yields an estimate of the geologic strike of the anomalous sources responsible for the observed geophysical anomalies. Therefore, we can obtain an improved strike and depth estimation of the magnetic sources. Although the algorithm can filter and cluster any Euler data set, we recommend obtaining the best solutions possible before any clustering. Hence, we have used a hybrid 3D extended Euler and 3D Werner deconvolution algorithm. We have developed synthetic and real examples from the Bathurst Mining Camp (New Brunswick, Canada). The output of this algorithm can be used as an input to any 3D geologic-modeling package.