CNES/GRGS 10-day gravity field models (release 2) and their evaluation

GRACE, designed to monitor temporal variations in the fluid mass at the surface of the Earth, is still operating and providing invaluable data 7 years after launch. One hundred and ninety-nine satellite-only geopotential solutions to degree and order 50 were recomputed per 10-day interval for the period 29 July 2002–27 May 2008 using an improved data editing and solution regularization procedure. These release 2 solutions are significantly improved compared to release 1 solutions, the noise over deserts and oceans in the form of North–South striping being reduced by 20–40%. This is thanks to the tailored regularization of each individual Stokes coefficient applied in the solution procedure, and to a time-variable reference model containing mean annual, semiannual and secular variations for degrees 2–50 towards which the variations per 10-day solution are constrained. It may attenuate signals of the order of a few percent, whereas this always occurs when applying a Gaussian smoother even with a half-width smoothing radius as small as 300 km. The uncertainty of an individual point in the time series of a basin expressed in equivalent water height inferred from the 10-day solutions is approximately 20 mm. Comparison of these 10-day solutions to monthly GRACE project solutions (CSR, GFZ and JPL) shows substantial differences. Even for the largest basin, the Amazon, a 15% difference in annual amplitude is found between CNES release 2 and CSR versus GFZ and JPL. The mass-loss estimates for East and West Greenland vary by 100%. Sometimes clear outliers are detected in the GFZ and JPL solutions when a particular basin is studied, which have to be eliminated. In view of the large differences detected between the time series for specific basins, it is hazardous to draw conclusions based on a single solution.