Signal and error in mass change inferences from GRACE: the case of Antarctica

This paper examines mass changes from Gravity Recovery and Climate Experiment (GRACE) monthly spherical harmonic gravity field solutions, focusing on the application to the Antarctic ice sheet. Regional integration approaches are investigated and compared with an alternative technique of fitting predefined mass change patterns. The main thrust of this new analysis is to better define error sources and error bounds on mass changes of the Antarctic ice sheet. Toward this purpose, we examine the release-4 GRACE products by the GeoForschungsZentrum Potsdam. Although errors related to the correction for glacial isostatic adjustment are significant, both leakage and GRACE error effects are also important. In particular, the details of how mass change interior to the Antarctic continent is weighted and scaled is crucially linked to the end-result mass trend solution. Accounting for error correlations in the GRACE monthly solutions, it is shown that the corresponding errors in Antarctic mass changes are about two times larger than predicted by uncorrelated error propagation. Our ice mass trend estimate for the grounded ice sheet in the period from 08/2002 to 01/2008 is (-109 +/- 48) Gt yr(-1), equivalent to a mean eustatic sea level rise of (+0.30 +/- 0.13) mm yr(-1). We also provide estimates for 16 individual drainage basins. Mass losses concentrate in the coastal regions of West Antarctica and on the Antarctic Peninsula with together (-105 +/- 35) Gt yr(-1). The Amundsen Sea Sector alone contributes (-60 +/- 8) Gt yr(-1). Although these findings agree with independent observations, GRACE trends deduced from a short time-series come with the cautioning footnote that the strength of interannual variations remains undetermined. We finally discuss directions of methodological improvements.