Page - 116 - 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

identified, most prominent amongst them the spike on June 13, 2010. The first and last increase in power can however not be identified in the plot any more. These increases in power are prominently visible in the lower panel of fig. 8.11, which shows the power of the residuals due to the AOC uncertainty eˆAOC (blue). The increase in power over a large range of frequencies during the long period of large opening angles starting on June 17, 2010 is now clearly identifiable here. Indeed, the increase in power is not limited to the frequency band identified previously (horizontal grey lines), but can also be observed at frequencies above 20mHz. KBR system noise is dominant in this frequency band and obscures this signal during normal operations. The PSD for the non-stationary AOC contribution (blue in fig. 8.10, also right margin of bottom panel in fig. 8.11) actually shows a consistent decrease in power with increased frequency, illustrating the unsuitability of the tool “PSD” to study this signal. A similar increase in power over a wide frequency spectrum can be observed in the peak on June 2, 2010, which is clearly visible in the AOC contribution to the residuals. As opposed to the period of increased opening angle later in the month, which is noted in the SDS monthly report for June 2010 (Flechtner, 2010), there is no indication of anomalous observations given for this time period. Possible sources of the increased AOC residual power at this time could also be found not in the observation geometry but in the SCA/ACC sensor fusion covariance matrices. Larger than normal SCA noise due to e.g. increased blinding could lead to reduced accuracies in the orientation estimate and in turn the AOC. Similarly, an origin in the angular rate observations could also play a role. Figure 8.12 shows the time-frequency-power diagram for the AOC contribution during the first days of June 2010 in the upper panel. The lower panel shows the opening angle β for GRACE-A and GRACE-B. The increase in power on June 2, 2010 corre- sponds roughly to some attitude excursions for GRACE-B. In general, the pointing for GRACE-B seems to undergo larger variations during this time span. For these two days, peaks in the opening angle could possibly be interpreted to correspond roughly to increased power in the residuals, although such a generalization should only be made cautiouslyand besubjected to further analysis.This ishowevera further indicationthat the AOC covariance matrices to some extent represent the true stochastic behaviour of the AOC, and that they are a valuable addition to the stochastic model for processing GRACE data. 8.4 Discussion This chapter focused on investigating the effects of uncertainties in the satellite orienta- tion on the determination of the antenna offset correction used in correcting GRACE KBR ll-SST data. The impact of the orientation uncertainty on other uses of the satellite orientation, such as in rotating the on-board accelerometer observations to the CRF for force integration, were not studied. Chapter8 Star Camera Observations and Uncertainties116