Non-linear gravity modelling

Abstract

A non-linear optimisation procedure has been applied to the interpretation of two-dimensional gravity anomalies. In comparison with linear methods, the non-linear approach is more difficult to control but can give more realistic solutions. Successful application of the non-linear approach depends upon the selection of a reasonable starting model from rules of thumb. The model is then changed iteratively so that the difference between the calculated and observed values is minimised. The method of changing the model is dependent upon the optimisation procedure used. Three basic procedures are available: gradient descent methods, the simplex method and the Monte Carlo method. In the authors' experience, the simplex method is a relatively efficient optimisation procedure in this application.