Errors in Mass Balance estimates of Antarctica from ice mask and input-output inconsistencies, pinpointed by GRACE

Surface mass balance (SMB) is computed from regional climate models (RCM) using reanalysis data. Estimates of the SMB vary between RCMs due to differences such as the model set-up, physical parameterizations, and topography as well as ice mask. The ice mask in a model defines the surface covered by glacier ice. The differences in ice masks appear small, however it is here shown that it leads to important differences in SMB when integrated over the continent. To circumvent this area-dependent bias, intercomparison studies of modelled SMB use a common ice mask (Mottram et al., 2021). The SMB in areas outside the common ice mask is discarded. By comparing the native ice masks with the common ice mask used in Mottram et al. 2021 we find differences in integrated SMB of between 20.1 and 102.4 Gt per year over the grounded ice sheet when compared to the ensemble mean from Mottram et al. 2021. These differences are nearly equivalent to the entire Antarctic ice sheet mass imbalance identified in the IMBIE study.
SMB is particularly essential when estimating the total mass balance of an ice sheet via the input-output method, by subtracting ice discharge from the SMB to derive the mass change. We use the RCM HIRHAM5 to simulate the Antarctic climate and force a newly develop offline subsurface firn model, to simulate the Antarctic SMB from 1980 to 2017. We use discharge estimates from two previously published studies to calculate the regional scale mass budget. To validate the results from the input-output method, we compared the results to the gravimetry-derived mass balance from the GRACE/GRACE-FO mass loss time series, computed for the period 2002–2020. We find good agreement between the two input-output results and GRACE in West Antarctica, however, there are large disagreements between the two input-output methods in East Antarctica and over the Antarctic Peninsula. Over the entire grounded ice sheet, GRACE detects a mass loss of 900 Gt for the period 2002-2017, whereas the two input-output results show a mass gain of 500 Gt and a mass loss of 4000 Gt, depending on which discharge dataset is used. These results are integrated over the native HIRHAM5 ice mask. If we instead integrate over the common ice mask from Mottram et al. 2021, the results change from a mass gain of 500 Gt to a mass loss of 500 Gt, and a mass loss of 4000 Gt to a mass loss of 5000 Gt, over the grounded ice sheet for the period 2002-2017. While the differences in ice discharge remain the largest sources of uncertainty in the Antarctic ice sheet mass budget, our analysis shows that even a small area bias in ice mask can have a huge impact in high precipitation areas and therefore SMB estimates. We conclude there is a pressing need for a common ice mask protocol, to create an accurate harmonized updated ice mask.