Seite - 43 - 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

described by Klinger, 2018 making use of current release 2.0 level 1B data is presented in section 6.2. Initially, the GRACE level 1B data is converted to a GROOPS-internal data format. The data is then screened for outliers and time periods of reduced data quality, downsampled, and interpolated, as described by Klinger (2018). For further processing, the data is divided into continuous segments, or arcs. In ITSG- Grace2016, arcs of two different lengths are used. For integration of the force models and determination of the dynamic satellite orbit using variational equations, arcs of no shorter than 1h and no longer than 24h are created. These 24h-arcs are the variational arcs. For processing of ll-SST observations, arcs of no shorter than 1h and no longer than 3h are used, these are the short arcs. Both variational and short arcs are divided at midnight each day, so that no segment spans multiple days. Where data is not available for at least one continuous hour, it is discarded. After data preprocessing steps, the following data is available within variational arcs: Kinematic orbits ITSG-Grace2016 uses an in-house computed kinematic orbit, the precise orbit determination (POD) product (Zehentner, 2017; Zehentner and Mayer-Gu¨rr, 2016), computed from level 1B GPS data using the raw observation approach. In addition to spacecraft positions, this approach also gives a fully populated3Dcovariancematrixateachepoch,whichisusedinfurtherprocessing steps. Dynamic orbit The level 1B orbit solution delivered by the SDS is used as a priori information for the in-house dynamic orbit integration. Linear accelerations The level 1B linear accelerations are pre-calibrated using mod- elled accelerometer data (Klinger, 2018). Orientation product This is a result of level 1B star camera and accelerometer sensor fusion, as described by Klinger, 2018. In addition, the following data is available in short arcs: Kinematic orbits The same kinematic orbits that are used in the variational arcs are also kept at hand in short arcs. Orientation product The same is true for the orientation, which is required in the computation of the AOC, the stochastic modelling of the orientation uncertainty in chapter 8 and for the co-estimation of satellite orientations through a total least squares algorithm described in chapter 9. KBR observations As delivered by the SDS. Time of flight correction As delivered by the SDS. Antenna offset correction The AOC is computed using the improved orientation product from the SCA/ACC sensor fusion. In both the short arcs and the variational arcs, all data is now given with 5s sampling, except for the kinematic orbits and their 3D covariance matrices, which are given with 300s sampling. 6.1 Data Preprocessing 43