Global ocean mass (or barystatic sea level) is an important component of the sea level budget. It can be computed directly from the GRACE and GRACE-FO data, or it can be computed by summing contributions from the cryosphere and hydrosphere. Here, we present a sensitivity analysis of the direct computation, using both averaging of mascon solutions as well as avering kernel methods applied to the spherical harmonics. We will discuss sensitivity to various geocenter models and GIA models, as well as discuss whether scaling is required for the averaging kernel method. For the scaling tests, we will utilize a simulation that includes trend fingerprints from icesheets and glaciers. We also discuss the effect of using different averaging areas (global oceans, global oceans excluding Hudson Bay, oceans with a 300km mask).
We find consisent results between the spherical harmonic and mascon calculations within the estimated uncertainy of fits based on the residuals to the linear trend + annual sinuspid model. Trends will change by up to 0.2 mm/year if geocenter models based on GRACE and ocean moddel data are used, or up to 0.4 mm/year if SLR-based geocenter is used. Different versions of glacial isostatic models introduce trend differences of order 0.1-0.2 mm/year. These systematic errors should be considered in addition to uncertainty from least squares fits when assesing closure to the sea level budget.