ZARM thermospheric neutral density solution from GRACE accelerometer data: Approach, validation and comparison

 The neutral mass density of the upper thermosphere can be determined by orbit and accelerometer data from Low Earth Orbit (LEO) satellites. Especially the accelerometers of geodetic satellites, measuring the non-gravitational accelerations acting on these satellites, are a very useful observation for precise density estimation also on very short time scales. Currently, due to the lac of direct measurements, the most accurate atmospheric density estimates are computed from such data.

The density estimation is mainly based on three separate disciplines, which are: 1. Precise radiative non-gravitational force modeling, 2. Modeling of the interaction between the rarefied atmospheric gases and the satellite, i.e. modeling of drag coefficients, and 3. Calibration of the accelerometer data, usually by dynamic Precise Orbit Determination (POD).

Besides being the most accurate source for thermospheric density data, differences in published datasets are rather high. Depending on the temporal resolution and space weather conditions, differences between those datasets might range between 100% and 25% on very short time scales (tens of seconds) and longer time scales around orbit period (1.5 to 3 hours), respectively. The reason for these differences is often said to be the drag coefficient modeling, which is true for a prominent amount of the differences, but the other two main disciplines, the non-gravitational force modeling and the accelerometer calibration, distinctly add to the error budget, especially for low solar activity.

In this contribution we present our density estimation approach and compare our solution based on GRACE data to other published datasets. We show strengths and weaknesses of the different datasets and try to explain the reasons for the rather big differences. We show how different processing and modeling options influence the final solution. For the three main disciplines of the density estimation we have developed evaluation strategies to get a better insight of the overall error budget.