Abstract
We compute magnetic field models from mostly ground-based geomagnetic observations over the period 1956 to 2025. Our magnetic field modeling strategy employs secular variation estimates derived from annual differences of observatories monthly means. The resulting models are compared to a recent satellite-based geomagnetic field model (Chaos 7.16) that covers the entire satellite era from 2000 to 2025. Both models are setup to mimic the geomagnetic field variation by order 6 B-splines, which also provides robust results for the secular acceleration. While the core field descriptions of the two models highly agree (correlation $\sim$0.92 at spherical harmonic degree l = 14), the both descriptions of the secular variation deviates, most noticeable on intermediate length scale. This difference may be explained by the difference of source geometries within satellite and ground-based data sets. Satellite data consider every magnetic field generation below their orbital sphere to be of internal origin. However, this might be difficult as night-time and quiet-time ionospheric field generation may also map into these data as internal sources. An indication for this might be the higher power of ground-based secular variation models for spherical harmonic degrees l = 4 − 12. We also devise a way to derive temporally more detailed secular variation estimates. Here, we aim to identify short-term secular variation most vivid in the equatorial region and the imprint of a quasi-biennial variation that possibly originate in the ionosphere and/or magnetosphere. The presented model was used to deduce candidates for the different IGRF-14 derivatives (definitive field model for 2020, current field model for 2025, secular variation forecast for 2025-2030).