Page - 135 - 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

9.3.3 Effect on Antenna Offset Correction After each iteration of the TLS adjustment the estimated orientations are used to re-compute the antenna offset correction. This leads to a new improved antenna offset correction that is slightly different from the initial AOC computed from the SCA/ACC sensor fusion orientation. As the AOC is a non-stationary process, analysing its change in the spectral domain would be fallacious. Instead, the change of magnitude of the AOC at each epoch is analysed, which is not dependent on the absolute value of the AOC. For a month of regular data, e.g. April 2008, the mean magnitude of the AOC is≈3.8 ·10−8m/s. The mean of the estimated correction due to the improved TLS orientation estimate for the same month is≈3.5 ·10−9m/s, which is roughly one order of magnitude smaller. For other months, the magnitude of both the AOC and the correction can be significantly larger. Figure 9.5 shows the distribution of these deviations dAOC(τ)= AOCnew(τ) AOCold(τ) −1 (9.3.1) in percent. A value below 0% indicates that the new AOC for that epoch is smaller than the original, a value larger than 0% represents an increased AOC. As these deviations are computed in the range-rate domain, a translation of this change into a statement on the magnitude of the opening angle β is not directly possible. Instead, a smaller value for dAOC indicates a slower change in the opening angle, with a larger value representing a faster change. All curves are skewed very slightly towards negative values, indicating that the TLS estimate of the orientation tends to dampen high-velocity motion in the satellite pointing. −100 −75 −50 −25 0 25 50 75 100 Deviation in AOC dAOC due to ∆αˆ [%] 0.00 0.01 0.02 0.03 Two SCA heads active: Both spacecraft Neither spacecraft GRACE-A only GRACE-B only Figure 9.5: Distribution of change in AOC due to improved orientation in relation to the magnitude of the original AOC. The data sets underlying the separate distributions were selected according to how many SCA heads were active for the respective epochs. 9.3 Results 135