Insights into the ITSG-Grace2016 processing
19/09/2016 | 10:30 | Session 1: Data Processing strategies and new concepts
Author(s): Torsten Mayer-Gürr, Beate Klinger, Andreas Kvas, Norbert Zehentner, Matthias Ellmer and Saniya Behzadpour
Torsten Mayer-Gürr, Beate Klinger, Andreas Kvas, Norbert Zehentner, Matthias Ellmer and Saniya Behzadpour
The ITSG-Grace2016 gravity field solutions are the latest release computed at TU Graz. The release consists of monthly unconstrained and daily Kalman filtered solutions. Compared to the former ITSG-Grace2014 release, multiple improvements within the processing chain have been implemented: updated background models, instrument data screening and calibration, improved numerical orbit integration, and covariance function estimation.
One of the team's main efforts is stringent modelling of noise sources and influences. To this end, we analyse the spatial and spectral behaviour of instrument noise and scrutinize the residuals of our estimations. This allows us to identify limiting factors in the overall process, and to develop strategies to minimize their impact. A prominent example is our pioneering use of fused sensor data: We improve the attitude information provided through the star camera by combining it with angular accelerations from accelerometer measurements. The resulting product shows marked improvements in noise behaviour, and leads to improvements in our solutions.
Our investigations go beyond satellite-bound error sources. Inaccuracies in background models contribute a significant amount of noise towards the overall error budget. We mitigate this effect by co-estimating high-frequency temporal variations in the background models through constrained daily gravity field solutions.
In our processing chain, we employ sophisticated noise modelling to estimate covariance functions for our observables, which are used throughout the final adjustment process. To make this knowledge available to other researchers, we publish our gravity field solutions together with complete variance-covariance information.
Noise is not the only focus of our research. We further investigate numerical effects and their impact on the solutions, for example in the integration of the equation of motion.
This contribution will highlight and present some of the points discussed above.