Uncertainty estimates of a GRACE inversion modelling technique over Greenland using a simulation

The low spatial resolution of GRACE causes leakage, where signals in one location spread out into nearby regions. Because of this leakage, using simple techniques such as basin averages may result in an incorrect estimate of the true mass change in a region. A fairly simple least squares inversion technique can be used to more specifically localize mass changes into a pre-determined set of basins of uniform internal mass distribution. However, the accuracy of these higher resolution basin mass amplitudes has not been determined, nor is it known how the distribution of the chosen basins affects the results. We use a simple 'truth' model over Greenland as an example case, to estimate the uncertainties of this inversion method and expose those design parameters which may result in an incorrect high-resolution mass distribution. We determine that an appropriate level of smoothing (300-400 km) and process noise (0.30 cm(2) of water) gets the best results. The trends of the Greenland internal basins and Iceland can be reasonably estimated with this method, with average systematic errors of 3.5 cm yr(-1) per basin. The largest mass losses found from GRACE RL04 occur in the coastal northwest (-19.9 and -33.0 cm yr(-1)) and southeast (-24.2 and -27.9 cm yr(-1)), with small mass gains (+1.4 to +7.7 cm yr(-1)) found across the northern interior. Acceleration of mass change is measurable at the 95 per cent confidence level in four northwestern basins, but not elsewhere in Greenland. Due to an insufficiently detailed distribution of basins across internal Canada, the trend estimates of Baffin and Ellesmere Islands are expected to be incorrect due to systematic errors caused by the inversion technique.