GRACE gravimetry provides a direct measure of water storage changes over continents. Thus, this novel technique enables – for the first time – to close the continental water balance on catchments with gauged discharge. We propose to use GRACE gravimetry to directly determine large scale actual evapotranspiration ETa from terrestrial precipitation and discharge measurements.However, the anisotropic error structure of conventional GRACE products is limiting their utility even for the largest basins available. To this end, we develop a new approach to GRACE error modelling, that makes use of known hydrological mass changes over selected areas, e.g. with negligible inputs (deserts) or with negligible evapotranspiration aET (snow/ice, high latitude / altitude regions). Beyond such hydrological constraints we want to further improve the GRACE solutions by adding surface mass loading constraints, as derived from deformation analysis of global GPS networks. An independent way to close the continental water balance on gauged catchments is the use of vertical integrated atmospheric moisture flux divergences Q¿¿ corresponding to P–ETa. The two approaches allow the setup of a consistent frame work between GRACE, hydrological and hydrometeorological mass change estimates as the basis for a quantification of errors.Both approaches, the evapotranspiration regionalized from directly determined values on gauged basins together with GRACE-derived water storage changes /Mt¿¿as well as the ¿ ¿Q and dM/dt approach provide an independent way for the determination of discharge from ungauged basins. Regional atmospheric model approaches are expected to improve the accuracy for ¿ ¿Q calculations compared to global atmospheric models. Based on the two approaches the global runoff from landmasses will be determined with quantified errors.