Abstract:
The availability of high-precision geomagnetic measurements from satellites such as Orsted and CHAMP opens a new era in geomagnetic field research. However, in order to take full advantage of the improved data accuracy it is necessary to refine the usual way of deriving field models from satellite data. This paper describes the derivation of a spherical harmonic model of the main field (up to degree/order 29) and of the secular variation (up to degree/order 13) using Orsted data spanning more than 2 yr (1999 March-2001 September) and applying new modelling approaches for a correct statistical treatment of the data errors and for considering external field contributions. Magnetospheric contributions are modelled up to degree/order two; the zonal terms vary with annual and semi-annual periodicity, and terms with degree n = 1 are modulated with the strength of the magnetospheric ring current as measured simultaneously by globally distributed geomagnetic observatories. In addition, the observatory data are used to constrain secular variation. The model is estimated using an iteratively reweighted least-squares method with Huber weights to account for the non-Gaussian data error distribution. The rms misfit achieved at non-polar latitudes is 3 nT for the scalar intensity and for one of the vector components perpendicular to the magnetic field; the third vector component (rms misfit of 6.4 nT owing to attitude noise) is downweighted when estimating the model. Comparing model predictions with actual scalar magnetic field observations from the CHAMP satellite yields an rms misfit of 3.4 nT at non-polar latitudes and 5.4 nT at polar latitudes.