A comparison of the gravity field over Central Europe from superconducting gravimeters, GRACE and global hydrological models, using EOF analysis

We analyse data from seven superconducting gravimeter (SG) stations in Europe from 2002 to 2007 from the Global Geodynamics Project (GGP) and compare seasonal variations with data from GRACE and several global hydrological modelsGLDAS, WGHM and ERA-Interim. Our technique is empirical orthogonal function (EOF) decomposition of the fields that allows for the inherent incompatibility of length scales between ground and satellite observations. GGP stations below the ground surface pose a problem because part of the attraction from soil moisture comes from above the gravimeter, and this gives rise to a complex (mixed) gravity response. The first principle component (PC) of the EOF decomposition is the main indicator for comparing the fields, although for some of the series it accounts for only about 50 per cent of the variance reduction. PCs for GRACE solutions RL04 from CSR and GFZ are filtered with a cosine taper (degrees 2040) and a Gaussian window (350 km). Significant differences are evident between GRACE solutions from different groups and filters, though they all agree reasonably well with the global hydrological models for the predominantly seasonal signal. We estimate the first PC at 10-d sampling to be accurate to 1 mu Gal for GGP data, 1.5 mu Gal for GRACE data and 1 mu Gal between the three global hydrological models. Within these limits the CNES/GRGS solution and ground GGP data agree at the 79 per cent level, and better when the GGP solution is restricted to the three above-ground stations. The major limitation on the GGP side comes from the water mass distribution surrounding the underground instruments that leads to a complex gravity effect. To solve this we propose a method for correcting the SG residual gravity series for the effects of soil moisture above the station.