Applications of Variance Component Estimation in GRACE-FO and SLR Spherical Geodetic Satellite Data Processing for Gravity Field Determination

Since mid-2018, the satellite pair of the gravimetry satellite mission called Gravity Recovery And Climate Experiment Follow-on (GRACE-FO) has been providing observations to determine the time-variable Earth’s gravity field with high temporal and spatial resolution. However, some of the low-degree Earth’s gravity field coefficients can better be determined by using Satellite Laser Ranging (SLR) observations to spherical geodetic satellites. Consequently, individual Earth’s gravity field coefficients may be replaced in the GRACE-FO products, or alternatively, the parameter estimation of the Earth’s gravity field can be performed using multi-technique combinations.
The Variance Component Estimation (VCE) is a well-established and widely adopted technique in satellite geodesy. As an example, VCE is used for combining observations from different geodetic space techniques or even to or from individual satellites to estimate precise geodetic parameters, e.g., to derive an international terrestrial reference frame.
In this study, applications of the VCE method, implemented in a development version of the Bernese GNSS software, are assessed for improved and more automated gravity field parameter determination. The primary focus is on the use of the VCE method for data quality assessments and to strengthen the orbit parametrization. In GRACE-FO data processing, reduced-dynamic orbits, where the usual orbit parameter set is extended with regular Piece-Wise Accelerations (PCAs), are used. In case of SLR data, the strong correlation between the gravity field coefficients and some of the dynamic orbit parameters prevents certain orbit parameters from being estimated but to compensate for this, stochastic pulses are set up. Both PCAs and the stochastic pulses need to be constrained in a way, that they can account for mis-modelings but preserve the sensitivity to the gravity field signal. It is shown that these constraints can be computed directly from VCE rather than determined empirically.