Satellite Magnetic Residuals Investigated With Geostatistical Methods

The geomagnetic field varies on a variety of time- and length scales, which are only rudimentarily considered in most present field models. The part of the observed field that cannot be explained by a given model, the model residuals, is often considered as an estimate of the data uncertainty (which consists of measurement errors and unmodeled signal), and is typically assumed to be uncorrelated and Gaussian distributed. We have applied geostatistical methods to analyze the residuals of the Oersted (09d/04) field model (www.dsri.dk/Oersted/Field models/IGRF 2005 candidates/), which is based on 5 years of Ørsted and CHAMP data and includes secular variation and acceleration, as well as low-degree external (magnetospheric) and induced fields. The analysis is done in order to find the statistical behavior of the space-time structure of the residuals, as a proxy for the data covariances. Once these are known, they can be used in the modeling process to improve the model, estimate its uncertainty, and eventually determine the optimal sampling rate in time and space. The uncertainty estimate on the field model can be used to estimate the uncertainty on quantities deduced from the field model, when the field model is used for geophysical interpretations.