GRACE gravity field recovery with background model uncertainties

In this article we present a computationally efficient method to incorporate background
model uncertainties into the gravity field recovery process. While the geophysical
models typically used during the processing of GRACE data, such as the atmosphere
and ocean dealiasing product, have been greatly improved over the last years, they
are still a limiting factor of the overall solution quality. Our idea is to use information
about the uncertainty of these models to find a more appropriate stochastic model for
the GRACE observations within the least squares adjustment, thus potentially
improving the gravity field estimates. We used the ESA Earth System Model to derive
uncertainty estimates for the atmosphere and ocean dealiasing product in the form of
an autoregressive model. To assess our approach, we computed time series of
monthly GRACE solutions from L1B data in the time span of 2005 to 2010 with and
without the derived error model. Intercomparisons between these time series show that
noise is reduced on all spatial scales, with up to 25\% RMS reduction for Gaussian
filter radii from 250 km to 300 km, while preserving the monthly signal. We further
observe a better agreement between formal and empirical errors, which supports our
conclusion that used uncertainty information does improve the stochastic description of
the GRACE observables.