[1] Redistribution of air masses due to atmospheric circulation causes loading deformation of the Earth's crust, which can be as large as 20 mm for the vertical component and 3 mm for horizontal components. Rigorous computation of site displacements caused by pressure loading requires knowledge of the surface pressure field over the entire Earth surface. A procedure for computing three-dimensional displacements of geodetic sites of interest using a 6 hourly pressure field from the National Centers for Environmental Prediction numerical weather models and the Ponte and Ray [2002] model of atmospheric tides is presented. We investigated possible error sources and found that the errors of our pressure loading time series are below the 15% level. We validated our model by estimating the admittance factors of the pressure loading time series using a data set of 3.5 million very long baseline interferometry observations from 1980 to 2002. The admittance factors averaged over all sites are 0.95 ± 0.02 for the vertical displacement and 1.00 ± 0.07 for the horizontal displacements. For the first time, horizontal displacements caused by atmospheric pressure loading have been detected. The closeness of these admittance factors to unity allows us to conclude that on average, our model quantitatively agrees with the observations within the error budget of the model. At the same time we found that the model is not accurate for several stations that are near a coast or in mountain regions. We conclude that our model is suitable for routine data reduction of space geodesy observations.

S U M M A R YThis paper is devoted to the simultaneous determination of the coseismic and postseismic gravitational changes caused by the great 2004 December 26 Sumatra-Andaman earthquake from the time-variable global gravity fields recovered by the Gravity Recovery And Climate Experiment (GRACE) mission. Furthermore, a complete modelling of the elasto-gravitational response of a self-gravitating, spherically layered, elastic earth model is carried out using a normal-modes summation for comparison with the observed coseismic gravitational change. Special attention is paid to the ocean mass redistribution. Special care is paid during the inversion of the data to avoid contamination of tectonic gravity changes by ocean tidal model errors, seasonal and interannual signals originating from continental hydrology and oceanic circulation as well as contamination of the coseismic gravity change by the postseismic relaxation. We use a 4.6-yr-long time-series of global gravity solutions including 26 months of postseismic data, provided by the Groupe de Recherche en Géodésie Spatiale (GRGS). For comparison, the Release-04 solutions of the Center for Space Research (CSR) are also investigated after a spectral windowing or a Gaussian spatial smoothing. Results are shown both in terms of geoid height changes and gravity variations. Coseismic and postseismic gravitational changes estimated from the different gravity solutions are globally similar, although their spatial extent and amplitude depend on the type of filter used in the processing of GRACE fields. The highest signal-to-noise ratio is found with the GRGS solutions. The postseismic signature has a spectral content closer to the GRACE bandwidth than the coseismic signature and is therefore better detected by GRACE. The coseismic signature consists mainly of a strong gravity decrease east of the Sunda trench, in the Andaman Sea. A gravity increase is also detected at a smaller scale, west of the trench. The model for the coseismic gravity changes agrees well with the coseismic signature estimated from GRACE, regarding the overall shape and orientation, location with respect to the trench and order of magnitude. Coseismic gravity changes are followed by a postseismic relaxation that are well fitted by an increasing exponential function with a mean relaxation time of 0.7 yr. The total postseismic gravity change consists of a large-scale positive anomaly centred above the trench and extending over 15 • of latitude along the subduction. After 26 months, the coseismic gravity decrease has been partly compensated by the postseismic relaxation, but a negative anomaly still remains south of Phuket. A dominant gravity increase extends over 15 • of latitude west of the trench, being maximal south of the epicentre area. By investigating analyses of two global hydrology models and one ocean general circulation model, we show that our GRACE estimates of the coseismic and postseismic gravitational changes are almost not biased by interannual variations originating from continental hyd...

[1] The differences between mass concentration (mascon) parameters and standard Stokes coefficient parameters in the recovery of gravity information from gravity recovery and climate experiment (GRACE) intersatellite K-band range rate data are investigated. First, mascons are decomposed into their Stokes coefficient representations to gauge the range of solutions available using each of the two types of parameters. Next, a direct comparison is made between two time series of unconstrained gravity solutions, one based on a set of global equal area mascon parameters (equivalent to 4°Â 4°at the equator), and the other based on standard Stokes coefficients with each time series using the same fundamental processing of the GRACE tracking data. It is shown that in unconstrained solutions, the type of gravity parameter being estimated does not qualitatively affect the estimated gravity field. It is also shown that many of the differences in mass flux derivations from GRACE gravity solutions arise from the type of smoothing being used and that the type of smoothing that can be embedded in mascon solutions has distinct advantages over postsolution smoothing. Finally, a 1 year time series based on global 2°equal area mascons estimated every 10 days is presented.