Assessment of atmospheric reductions for terrestrial gravity observations

Atmospheric attraction and loading effects account for about 10% of all observed time-dependent gravity variations in which the dominant gravity signal is the tides of the solid Earth. The impact can be roughly estimated using barometric pressure from the observation site, but in that case only atmospheric variations which are correlated with local barometric pressure changes are taken into account. If e.g. geodynamic signals are to be investigated, the variations which are unconsidered have an amplitude of several μGal, therefore they are large enough to require consideration. In this paper, three different procedures to remove the atmospheric effect are compared. Numerical results show that the atmospheric reduction using threedimensional (3D) data from the European Center for Medium-Range Weather Forecasts (ECMWF) ought to be computed up to 5° around a station. A peak-topeak amplitude of the differences between a reduction using 3D data and one using two-dimensional (2D) data from the ECMWF is 0.5 μGal including seasonal variations with an amplitude of 0.15 μGal, and it has a Root Mean Square (RMS) value of 0.1 μGal considering a time span of 4 years. The amplitude of reduction based on a regression coefficient/admittance factor differs from the two physical methods by approximately 3 μGal with a RMS value of 0.4 μGal in the same 4 year-long observation period. From spectral analyses of the three reductions it emerges that the amplitudes of the more comprehensive methods are 11-12% smaller than the reduction using an admittance factor in the spectral range from 0.0 CPD (cycle per day) to 0.18 CPD on average. Investigation of the atmospheric reduction effects on the tidal analysis indicates that there are visible improvements in the tidal analysis using the reductions based on the physical approaches compared to the reduction using an admittance factor but not between the two physical approaches. Concerning the amplitude factor of the polar motion signal, there is 1.5% difference in the value of the factor after applying the two physical methods. The differences between the physical approaches stem from the consideration of the air density distribution. A peak-to-peak amplitude is about 0.5 μGal when the attraction effect up to 5° around a station is computed. The omission of vertical variations in air density leads to inaccuracies which should be avoided, for instance, in the validation of non-tidal ocean loading effects or studies of tectonic phenomena.