The geomagnetic jerk of 2003.5-characterisation with regional observatory secular variation data

Abstract

The 2003.5 geomagnetic jerk was identified in geomagnetic records from satellite data, and a matching feature reported in variations in length-of-day (ΔLOD), but detailed study has been hampered by lack of geomagnetic observatory data where it appears strongest. Here we examine secular variation (annual differences of monthly means) based on a new resource of 43 Chinese observatory records for 1998 until the present, focusing on 10 series of particularly high quality and consistency. To obtain a clean series, we calculate the covariance matrix of residuals between measurements and a state-of-the-art field model, CHAOS-6, and use eigenvalue analysis to remove noisy contributions from the uncorrected data. The magnitude of the most significant eigenvector correlates well with Dcx (corrected, extended Dst), suggesting the noise originates from unmodelled external magnetic field. Removal of this noise eliminates much coherent misfit around 2003—2005; nevertheless, the 2003.5 jerk is seen clearly in the first time derivative of the East component in Chinese data, and is also seen in the first time derivative of the vertical component in European data. Estimates of the jerk time are centred on 2003.5, but with some spatial variation; this variation can be eliminated if we allow a discontinuity in the secular variation as well as its temporal gradient. Both regions also provide evidence for a jerk around 2014, although less clearly than 2003.5. We create a new field model based on new data and CHAOS-6 to further examine the regional signals. The new model is close to CHAOS-6, but better fits Chinese data, although modelling also identifies some data features as unphysical.