Can core‐surface flow models be used to improve the forecast of the Earth's main magnetic field?

Geomagnetic main field models used for navigation are updated every 5 years and contain a forecast of the geomagnetic secular variation for the upcoming epoch. Forecasting secular variation is a difficult task. The change of the main magnetic field is thought to be principally due to advection of the field by flow at the surface of the outer core on short timescales and when large length scales are considered. With accurate secular variation (SV) and secular acceleration (SA) models now available from new satellite missions, inverting for the flow and advecting it forward could lead to a more accurate prediction of the main field. However, this scheme faces two significant challenges. The first arises from the truncation of the observable main field at spherical harmonic degree 13. This can however be handled if the true core flow is large scale and has a rapidly decaying energy spectrum. The second is that even at a given single epoch the instantaneous SV and SA cannot simultaneously be explained by a steady flow. Nevertheless, we find that it may be feasible to use flow models for an improved temporal extrapolation of the main field. A medium‐term (≈10 years) hindcast of the field using a steady flow model outperforms the usual extrapolation using the presently observed SV and SA. On the other hand, our accelerated, toroidal flow model, which explains a larger portion of the observed average SA over the 2000–2005 period, fails to improve both the short‐term and medium‐term hindcasts of the field. This somewhat paradoxical result is related to the occurrence of so‐called geomagnetic jerks, the still poorly known dynamical nature of which remains the main obstacle to improved geomagnetic field forecasts.