Grace Time-Variable Gravity Accuracy Assessment

This study aims at the calibration of the formal variance covariance matrices of monthly GRACE-only gravity models that are known to give too optimistic accuracy measures for derived gravity functionals respectively surface mass anomalies. Based on 16 monthly solutions generated at GFZ Potsdam, a simple degree-dependent scaling of given variance-covariance matrices is considered, using difference degree amplitudes of monthly GRACE-only gravity field models that are separated by 12 months. It turns out that for the recent GFZ models the mean accuracy level of the monthly fields is about 20 - 40 times lower than the GRACE baseline accuracy. An observed scatter of the accuracy of the monthly models can be attributed to the amount of data used to generate the monthly model, but a more significant contribution arises from variations in the ground track coverage from month to month.Error propagation based on variance-covariance matrices calibrated with respect to 20 times the GRACE baseline, indicates that on global average the level of the model uncertainty is well reproduced when compared to residual surface mass variability in the space domain. Comparisons between time series of basin averages of surface mass variations from GRACE and state-of-the-art hydrology models seem to confirm this calibration factor for river basins with large annual mass variations. For basins with small variations the results suggest that a calibration with respect to 40 times the baseline may be a more representative value, however. As a general result it is observed that the formal respectively calibrated variance-covariance matrices of the monthly gravity models do not represent additional spurious gravity signals of significant amplitude caused by aliasing effects.