| Abstract: | The performances of least-squares orthogonal polynomial and relaxation techniques in the separation of regional and residual anomalies have been evaluated with a view to minimizing personal biasing. The advantage of orthogonal over nonorthogonal polynomials is their ability to estimate an optimum order of polynomial to represent the predominant regional trend in the data using an approximate 2D difference table, the Z-matrix. The correlation coefficients between residuals of two consecutive orders also give the same result. In the relaxation technique, a linear trend is assumed within each cell of the mesh of a square grid. A set of such linear segments can approximate any complicated regional trend. The performances of these two techniques have been evaluated using simulated gravity anomalies produced by 2D and 3D complex regional structures superimposed on residual fields due to cylinders and prismatic bodies, as well as three field examples taken from the published literature. The analyses have revealed that the relaxation technique produces excellent results when an optimum polynomial order rather than an arbitrary fixed one is used for computing the boundary conditions along the periphery of the map. Analyses have revealed that such boundary conditions provide minimum distortion near the two ends of the profile. |
