Quantifying the error level in computed magnetic amplitude data for 3D magnetization inversion

Abstract

Three-dimensional inversion plays an important role in the quantitative interpretation of magnetic data in exploration problems, and magnetic amplitude data can be an effective tool in cases in which remanently magnetized materials are present. Because amplitude data are typically calculated from total-field anomaly data, the error levels must be characterized for inversions. Lack of knowledge of the error in amplitude data hinders the ability to properly estimate the data misfit associated with an inverse model and, therefore, the selection of the appropriate regularization parameter for a final model. To overcome these challenges, we have investigated the propagation of errors from total-field anomaly to amplitude data. Using parametric bootstrapping, we find that the standard deviation of the noise in amplitude data is approximately equal to that of the noise in total-field anomaly data when the amplitude data are derived from the conversion of total-field data to three orthogonal components. We then illustrate how the equivalent source method can be used to estimate the error in total-field anomaly data when needed. The obtained noise estimate can be applied to amplitude inversion to recover an optimal inverse model by applying the discrepancy principle. We test this method on synthetic and field data and determine its effectiveness.