Seasonal and inter-annual mass variability in Antarctica from GRACE

The seasonal and inter-annual variability in Antarctic surface mass balance represent important components of the temporal variability in total ice sheet mass balance. At present, large uncertainties remain in our knowledge of those temporal components, spatial distribution and long-term trend. GRACE measurements of time variable gravity represent one of the only ways to directly measure the monthly ice sheet mass changes and - if accurate enough - provide important insights to constrain the seasonal and inter-annual variability in surface mass balance, which in turn is essential to evaluate global and regional atmospheric climate model outputs. In Antarctica, errors in the atmospheric mass correction applied to the GRACE data are the dominant uncertainties on the retrieval of seasonal and inter-annual mass changes. We estimate the error in monthly GRACE correction for the atmospheric mass distribution by comparing surface pressure fields from available GPS-RO (e.g. CHAMP, SAC-C, and COSMIC) and ECMWF pressure fields. We evaluate the results using available surface pressure measurements from Automatic Weather Stations (AWS). We compare the spatial pattern in our atmospheric error estimates with different estimates of surface mass balance (SMB) derived from regional atmospheric climate model outputs, from the continental scale down to the regional scale, to determine if those errors are associated to particular regional conditions and climatic regimes. We show that when averaged over the entire ice sheet or East or West Antarctica, the accuracy of the GRACE data is sufficient to constrain seasonal and inter-annual changes in surface mass balance. But when moving to smaller scales, e.g. at the scale of a drainage basin, we identify regions where the uncertainty in atmospheric mass correction is large enough to impair the scientific interpretability of the GRACE results. In those areas, we generate preliminary corrections to the GRACE data and evaluate the results.