Noise variance estimation and optimal weight determination for GOCE gravity recovery

In the course of level 2 data processing for the GOCE (Gravity Field and Steady–State Ocean Circulation Explorer) satellite mission different streams of level 1b data will be merged in a single solution providing the Earth’s gravity field, but also state-vector parameters and other quantities. A proper weighting of orbit tracking data, gravity gradiometry data and certain a priori information, usually considered as ‘solution constraints’, can be expected as crucial for the solution quality. But the a priori stochastic models, based on pre–mission assessment of the expected instrument behaviour, may be over–optimistic or even too pessimistic since they refer to an unprecedented mission with scientific payload never tested in space. One way to derive an optimal weighting scheme on a statistically sound basis while simultaneously validating the stochastic noise levels of the data is by including variance component estimation as a part of the level 1b to level 2 data analysis process. The idea is that by applying Monte-Carlo techniques this method can be extended to a large-scale problem like GOCE data analysis, using software modules that already exist or are currently under development. The proposed method has been tested using simulated GOCE orbit trajectories as well as gravity gradiometry data corrupted by colored random noise.