Combined determination of orbit and accelerometer calibration parameters for CHAMP

Highest quality reduced-dynamic orbits for CHAMP may be obtained without any use of additional accelerometer data. In this case a considerably large number of so-called pseudo-stochastic parameters (e.g., piecewise constant or piecewise linear accelerations) compensate for deficiencies in the dynamic model. The excellent agreement between such estimated accelerations and measured accelerations from the STAR accelerometer rises the question how a combined processing of GPS and accelerometer data should be optimally realized and which benefits could be expected from such a procedure. Two different approaches are studied: 1: Accelerometer measurements are used to define an empirical function (e.g., piecewise constant, piecewise linear, smoothed, ...) of the non-gravitational accelerations. By introducing this information into the orbit determination process only a small set of calibration parameters has to be estimated. 2: Accelerometer data are treated as additional observations to the GPS observations in the least squares parameter estimation process. This approach promises a statistically correct combination of both measurement types with a piecewise constant or piecewise linear acceleration function derived from these observations. Compared to the standard GPS-only reduced-dynamic orbit determination, accelerometer calibration parameters have to be determined again in addition to the stochastic acceleration parameters. For both approaches the interpretation of the estimated parameters will be analyzed and discussed. The constraining of stochastic parameters in the second approach is of particular interest.