Kalman filter based modeling of the Holocene geomagnetic field

Investigations of the Earth's internal magnetic field over millennial timescales are based on paleomagnetic data of thermoremanent or sedimentary origin. Both sources are affected by uneven spatial coverage, measurement errors, and dating uncertainties. In recent years, several Bayesian models have been developed to reconstruct the Holocene geomagnetic field, aiming to address these challenges and reflect the resulting uncertainties in the posterior distribution. Many of these approaches can be unified in the Gaussian process framework. Variations in assumptions about the magnetic field are reflected in the choice of priors, while differences in inversion strategies result in distinct posterior approximations.

 

We provide a brief overview of existing models and describe our approach in more detail, focusing on approximating the posterior using a Kalman filter. We discuss the selection of prior parameters and the consequences of different choices, leading to an update of the ArchKalmag model. The revised model's local predictions and global characteristics are presented, and our results are compared with existing models, with a particular focus on model uncertainties.