Mitigation of ionospheric signatures in Swarm GPS gravity field estimation using weighting strategies

Abstract

Even though ESA's three-satellite low-earth orbit (LEO) mission Swarm is primarily a magnetic field mission, it can also serve as a gravity field mission. Located in a near-polar orbit with initial altitudes of 480 km for Swarm A and Swarm C and 530 km for Swarm B and equipped with geodetic-type dual frequency Global Positioning System (GPS) receivers, it is suitable for gravity field computation. Of course, the Swarm GPS-only gravity fields cannot compete with the gravity fields derived from the ultra-precise Gravity Recovery And Climate Experiment (GRACE) K-band measurements. But for various reasons like the end of the GRACE mission in October 2017, data gaps in the previous months due to battery aging, and the gap between GRACE and the recently launched GRACE Follow-On mission, Swarm gravity fields became important to maintain a continuous time series and to bridge the gap between the two dedicated gravity missions. By comparing the gravity fields derived from Swarm kinematic positions to the GRACE gravity fields, systematic errors have been observed in the Swarm results, especially around the geomagnetic equator. These errors are already visible in the kinematic positions as spikes up to a few centimeters, from where they propagate into the gravity field solutions. We investigate these systematic errors by analyzing the geometry-free linear combination of the GPS carrier-phase observations and its time derivatives using a combination of a Gaussian filter and a Savitzky–Golay filter and the Rate of Total Electron Content (TEC) Index (ROTI). Based on this, we present different weighting schemes and investigate their impact on the gravity field solutions in order to assess the success of different mitigation strategies. We will show that a combination of a derivative-based weighting approach with a ROTI-based weighting approach is capable of reducing the geoid rms from 21.6 to 12.0 mm for a heavily affected month and that almost 10 % more kinematic positions can be preserved compared to a derivative-based screening.