Tailored accelerometer calibration by POD for thermospheric density computation with GRACE and GRACE-FO

The density of the upper atmosphere can be determined by orbit and accelerometer data from low Earth orbit satellites. Especially the accelerometers of geodetic satellites, measuring the non-gravitational accelerations acting on them are a very viable observation.

The density estimation is mainly based on three separate disciplines, which are: (1) Precise radiative non-gravitational force modelling, (3) Modelling of the interaction between the rarefied atmospheric gases and the satellite, i.e. modelling of drag coefficients, and (3) the calibration of the accelerometer data by dynamic Precise Orbit Determination (POD). This contribution focuses mainly on the last point. Nevertheless, we also validate the modelled radiative accelerations and use them as reference for the calibration results.

The accelerometers of all geodetic satellites are affected by a drifting bias and scale factors unequal to one. Therefore a calibration of the data is indispensable. Usually time dependent bias and scale factors are estimated. For standard POD or Gravity Field Recovery (GFR) these parameters are estimated together with empiric and other model parameters. In both cases, the estimated accelerometer calibration parameters are not of major interest, but improve the orbit fit or gravitational field coefficients. The used parametrizations and weighting strategies of the observation data, do not give realistic or physical accelerometer calibration results because parameters are not sensitive and effects are absorbed or smeared into other parameters and models. This is unsatisfying, especially for the anticipated use for the density determination.

In this contribution we use dynamic POD and investigate different parametrization strategies tailored for a physical accelerometer calibration. We investigated the effect of constraining the accelerometer calibration parameters in that way, that a continuous calibration over all arcs is achieved, where normally each arc is treated locally separated from all other arcs. The scale factor is concurrently estimated, but over a longer batch of arcs. We varied the length between one day and years. Furthermore, different calibration equations, different observation data and combinations (kinematic positions, science orbit data, inter-satellite ranging), weighting strategies, initial parameters and pre-processing of the accelerometer data is investigated. The validation of the results is not easy, because usual metrics like post-fit residuals do not reflect the quality of the accelerometer calibration. We introduce a validation approach using the modelled non-gravitational forces and also show the influence of the different calibration options on the resulting density or drag acceleration.

All results and data for the whole GRACE and GRACE-FO missions are available on our data server (link in presentation/ uploaded material).