Combination of GRACE monthly gravity fields on the normal equation level

A large number of time-series of monthly gravity fields derived from GRACE data provide users with a wealth of information on mass transport processes in the system Earth. The users are, however, left alone with the decision which time-series to analyze. Following the example of other well-known combination services provided by the geodetic community, the prototype of a combination service has been developed within the frame of the project EGSIEM (2015-2017) to combine the different time-series with the goal to provide a unique and superior product to the user community. Four associated analysis centers (ACs) of EGSIEM, namely AIUB, GFZ, GRGS and IfG, generated monthly gravity fields which were then combined using the different normal equations (NEQs). But the relative weights determined by variance component estimation (VCE) on the NEQ level do not lead to an optimal combined product due to the different processing strategies applied by the individual ACs. We therefore resort to VCE on the solution level to derive relative weights that are representative of the noise levels of the individual solutions. These weights are then applied in the combination on the NEQ level. Prior to combination, empirical scaling factors that are based on pairwise combinations of NEQs are derived to balance the impact of the NEQs on the combined solution. We compare the processing approaches of the different ACs and introduce quality measures derived either from the differences w.r.t. the monthly means of the individual gravity fields or w.r.t. a deterministic signal model. After combination, the gravity fields are validated by comparison to the official GRACE SDS RL05 time-series and the individual contributions of the associated ACs in the spectral and the spatial domain. While the combined gravity fields are comparable in signal strength to the individual time-series, they stand out by their low noise level. In terms of noise, they are in 90 of all months as good or better than the best individual contribution from IfG and significantly less noisy than the official GRACE SDS RL05 time-series