Choices for temporal gravity field modeling for precision orbit determination of CryoSat-2

In this paper we review the precision orbit determination (POD) performance of the CryoSat-2 mission where we used all tracking data between June-2010 and Jan-2023; with station and beacon coordinates provided in the ITRF2020 reference system, we use a mean gravity model, and we use spacecraft specific models for modeling drag and radiation pressure. To model time variable gravity (TVG) we distinguish between two components, there is a short term oceanic and atmospheric part for which we use the AOD1B model; for the longer term part we employ GRACE and GRACE-FO monthly potential coefficient solutions. Our experience is that adding TVG information is not necessarily successful during POD, and that attention must be paid to the proper processing of the GRACE and GRACE-FO data. To demonstrate this property we define four runs where we gradually implement TVG information. An evaluation criterion is the level of POD tracking residuals, the level of the empirical accelerations, and a comparison to precision orbit ephemeris provided by the Centre National d'Etudes Spatiales (CNES). Unexplained empirical accelerations found during POD are on the level of 3 nm/s ² for the along-track component and 13 nm/s ² for the cross-track component. The laser residuals converge at approximately 1.02 cm and the Doppler residuals are on the level of 0.406 mm/s, the radial orbit difference to the CNES POE-F (Precision Orbit Ephemeris version F) orbits narrows to 6.5 mm. Tracking residuals are not evenly distributed for DORIS (Doppler Orbitography and Radiopositioning Integrated by Satellite) beacons, the South Atlantic Anomaly effect is for instance clearly visible in the first empirical orthogonal function EOF mode of monthly binned DORIS residuals. After consideration of all possible TVG approaches our conclusion is that 3 hourly AOD1B model fields result in a small but visible improvement. The addition of TVG from GRACE and GRACE-FO is implemented in two different ways from which we can select a version that does lead to a reduction in the Doppler tracking residuals and which does reduce the level of solved for empirical accelerations.