The GOCE mission provides a unique gravity gradient dataset, which is used by most state-of-the-art global gravity field models. The quality of the gravity gradients of previous data releases degraded during the mission lifetime, which was often attributed to the increasing drag forces towards the end of the mission due to the approaching solar maximum and also, starting in August 2012, a sequence of five orbit lowerings. An imperfect calibration of the gravity gradiometer data was often suspected as root cause for the degradation. We present the novel method for the calibration of the GOCE gravity gradiometer data that was used for the official reprocessing of the GOCE gravity gradients in 2018. It is based on a comprehensive calibration model that includes quadratic factors and angular acceleration couplings as new calibration parameters. The calibration parameters are estimated from star tracker angular rates, gravity gradients calculated from a GRACE gravity field model and the condition that the three gradiometer arms measure the same non-gravitational acceleration. In addition to the GOCE data generated during science mode operations, we also use the data collected during satellite shaking mode operations. We demonstrate that applying the new method removes systematic errors to a large extent and, consequently, yields gravity gradients of superior and constant quality compared to previous data releases.

We analyse the inter-boresight angles (IBA) measured by the star trackers on board the GOCE satellite and find that they exhibit small offsets of 7–9″ with respect to the ones calculated from the rotation of the star tracker reference frames to the satellite reference frame. Further, we find small variations in the offsets with a peak-to-peak amplitude of up to 8″, which correlate with variations of the star trackers’ temperatures. Motivated by these findings, we present a method for combining the attitude quaternions measured by two or more star trackers that includes an estimation of relative attitude offsets between star trackers as a linear function of temperature. The method was used to correct and combine the star tracker attitude quaternions within the reprocessing of GOCE data performed in 2018. We demonstrate that the IBA calculated from the corrected star tracker attitude quaternions show no significant offsets with respect to the reference frame information. Finally, we show that neglecting the star tracker attitude offsets in the processing would result in perturbations in the gravity gradients that are visible at frequencies below 2 mHz and have a magnitude of up to 90 mE. The presented method avoids such perturbations to a large extent.