An estimate of the errors of the IGRF/DGRF fields 1945–2000

The IGRF coefficients inevitably differ from the true values. Estimates are made of the their uncertainties by comparing IGRF and DGRF models with ones produced later. For simplicity, the uncertainties are summarized in terms of the corresponding root-mean-square vector uncertainty of the field at the Earth’s surface; these rms uncertainties vary from a few hundred to a few nanotesla. (It is assumed that the IGRF is meant to model the long-wavelength long-period field of internal origin, with no attempt to separate the long-wavelength fields of core and crustal origin; the models are meant for users interested in the field near and outside the Earth’s surface, not for core-field theoreticians.) So far we have rounded the main-field coefficients to 1 nT; this contributes an rms vector error of about 10 nT. If we do in fact get a succession of vector magnetic field satellites then we should reconsider this rounding level. Similarly, for future DGRF models we would probably be justified in extending the truncation from n = 10 to n = 12. On the other hand, the rounding of the secular variation coefficients to 0.1 nT could give a false impression of accuracy.