Ice sheet mass balance in Antarctica measured by GRACE and its uncertainty

As a critical component of the cryosphere, the Antarctic Ice Sheet (AIS) has strong connection with the sea level change and global climate change. Accurate quantification of the current spatial and temporal mass changes of AIS is very important to improve our understanding and prediction of its response and contribution to global change. The Gravity Recovery and Climate Experiment (GRACE) mission has provided new and useful observations to detect AIS mass balance since its launch in March 2002. There are significant differences among the GRACE estimates of the total mass change. The big difference is due in part to considerable uncertainty in the accuracy of glacial isostatic adjustment (GIA) signals, and also due to use of different time spans, different versions of GRACE products and different GRACE post-processing methods. Using 124 monthly GRACE gravity field solutions of Release 5 (RL05) produced at the Center for Space Research (CSR) of the University of Texas, Austin, spanning the interval from January of 2003 through December of 2013, the mass balance of AIS is estimated by two post-processing ways: the optimizing averaging kernel method ( also named VW) and the combined filter method (the first step is called P5M11 decorrelation filter to remove correlated noise by fitting and subtracting a fifth-order polynomial to even and odd coefficient pairs at spherical harmonic orders eleven and above, the second involves smoothing with a 250 km Gaussian filter). A detailed error analysis is provided including consideration of leakage-in, leakage-out, and errors in modeling mass variations of the atmosphere, ocean and GIA. In addition, a statistical model selection criterion is employed in computation of trends from mass variation time series, and the impact of K1 tidal alias is analyzed. The results reveal that during 2003-2013, the total mass of the ice sheet decreased significantly at change rates of -163 +/- 150, -129 +/- 41 and -81 +/- 27 Gt/a for three GIA models: GW13, IJ05, W12a. There was a distinct region with mass loss in the Amundsen Sea Embayment of West Antarctic ice sheet and the Northern Antarctic Peninsula, while an increasing mass gain was concentrated in the Dronning Maud Land and the Enderby Land of East Antarctic ice sheet. Furthermore, we use hypotheses testing and information criteria evaluation to select the best trend model fitting together with sinusoidal functions of annual (365. 0-d) and semi-annual (181. 0-d) signals and the S2 (161. 0-d), K1 (2725.4-d) and K2 (1362.7-d) tidal aliases. We found that K1 tidal alias has a potential to falsify the acceleration estimates. Although it is not good enough to confirm the K1 tidal alias based on an eleven-year time-series, the impact of K1 tidal alias deserves further notice. By comparing the quantities of total mass balance computed by the two different processing methods and three different GIA models in the Antarctica, we find that the differences are less than 15 Gt/a between two processing methods, but the largest difference is about 80 Gt/a between different GIA models. The analysis of the uncertainty of GRACE's estimation of AIS mass balance indicates that the largest source of error is the GIA correction. Our results indicate that during January 2003 to December 2013 the contribution of AIS to sea level rise was about +0.34 +/- 0.11 mm/a. Significant mass loss increases were limited to the basin that contains Pine Island Glacier along the Amundsen Sea coast of West Antarctica. During the analyzed time period, the total mass acceleration was -8 +/- 10 Gt/a(2), equivalent to +0.02 +/- 0.03 mm/a(2) sea level rise. Results of analysis point to the conclusion that when using a given GRACE data set with same error correction, the differences of total mass changes are not highly dependent on which post-processing strategies to be used but on the different GIA models. Therefore, a more accurate GIA model is the key for determining Antarctic ice mass change from GRACE in the present and future.