The research unit NEROGRAV: first results on stochastic modeling for gravity field determination with real GRACE and GRACE-FO data

The central hypothesis of the Research Unit (RU) NEROGRAV (New Refined Observations of Climate Change from Spaceborne Gravity Missions), funded for three years by the German Research Foundation DFG, reads: only by concurrently improving and better understanding of sensor data, background models, and processing strategies of satellite gravimetry, the resolution, accuracy, and long-term consistency of mass transport series from satellite gravimetry can be significantly increased; and only in that case the potential of future technological sensor developments can be fully exploited. Two of the individual projects (IPs) within the RU work on stochastic modeling for GRACE and GRACE-FO gravity field determination. TU München and TU Berlin are responsible for IP4 (OSTPAG: optimized space-time parameterization for GRACE and GRACE-FO data analysis), where besides optimal parameterization the focus is on the stochastic modeling of the key observations, i.e. GRACE and GRACE-FO inter-satellite ranging and accelerometer observations, in a simulation (TU München) and real data (TU Berlin) environment. IP5 (ISTORE: improved stochastic modeling in GRACE/GRACE-FO real data processing), which GFZ is responsible for, works on the optimal utilization of the stochastic properties of the main GRACE and GRACE-FO observation types and the main background models.

This presentation gives first insights into the TU Berlin and GFZ results of these two IPs which are both related on stochastic modeling for real data processing based on GFZ GRACE and GRACE-FO RL06 processing. We present the analyses of K-band inter-satellite range observations and corresponding residuals of three test years of GRACE and GRACE-FO real data in the time and frequency domain. Based on the residual analysis we show results of the effects of different filter matrices, which take into account the stochastic properties of the range observations in order to decorrelate them. The stochastic modeling of the background models starts with Monte-Carlo simulations on background model errors of atmospheric and oceanic mass variations. Different representations of variance-covariance matrices of this model information are tested as input for real GRACE data processing and their effect on gravity field determination are analyzed.