Page - 99 - in Contributions to GRACE Gravity Field Recovery - Improvements in Dynamic Orbit Integration, Stochastic Modelling of the Antenna Offset Correction, and Co-Estimation of Satellite Orientations

Star Camera Observations and Uncertainties 8 In the processing strategy employed for ITSG-Grace2016, an accurate stochastic model is indispensable to determining a high-quality gravity field solution. In the regime used to estimate the stochastic model, which was described in section 6.5, the complete noise spectrum for the ll-SST observations was modelled as resulting from exactly one stationary process. The estimated covariance function was subsequently scaled by an arc-wise variance factor. This scaling equally affects all time lags of the covariance func- tion, or equivalently all frequencies of the PSD. It does not change the assumption of stationarity within one arc, but only scales the variance of the stationary process. These arc-wise variance factors, in effect, can be regarded as a fudge factor for unmodelled variations in the observation noise. Analysis of the arc-wise variance factors for ITSG-Grace2016 has shown that they are at times correlated with the magnitude and change of the satellite pointing angles with respect to the line of sight frame. The magnitude of these angles maps directly into the magnitude of the antenna offset correction. As the satellites are subject to active steering and pointing variations due to environmental effects, the variance in the AOC over one month of observations, or even one arc, can decidedly not be regarded as the result of a stationary process. This chapter introduces an additional non-stationary stochastic model for the antenna offset correction. This new AOC stochastic model is derived from the full orientation covariance matrices obtained in the improved sensor fusion described in section 6.2. The impact of introducing this a priori information on the non-stationary AOC noise alongside the estimated stationary stochastic model (see section 6.5) is analysed. The combined stochastic models are used to estimate a time series of GRACE monthly gravity field solutions, based on the ITSG-Grace2016 processing chain. The focus in the analysis will not be on the gravity field solutions themselves, but on the stochastic model and post-fit residuals in the ll-SST observable. The impact on the estimated Stokescoefficientswillbeanalysedinchapter9, togetherwiththegravityfieldsolutions estimated therein. 8.1 The Antenna Offset Correction in the ll-SST Observation Equation The antenna offset correction is one of many corrections applied to the ll-SST KBR observations. The reduced observation vector for the low-low satellite-to-satellite 99