Treatment of ocean tide background model errors in the context of GRACE/GRACE-FO data processing

Due to its high-frequency, high-amplitude nature the ocean tide (OT) signal poses a core limitation within the GRACE/GARCE-FO data processing, as its temporal undersampling ultimately results in the typical striping pattern. To some degree this issue can be resolved by employing a-priori de-aliasing based on an OT background model. However, due to model errors and imperfections some residual effects inevitably remain within the resulting gravity product.

So far, the background models have been assumed as error-free within the data processing. Since ocean tide models feature distinct, time-invariant spatial error patterns (low uncertainties in open oceans, high uncertainties in coastal and high-latitude regions), it is reasonable to assume that weighting the observations based on the underlying spatial error shall result in a more homogenous gravity solution.

Thus, error co-variance matrices are derived for the spherical harmonic coefficients representing the eight principal tides based on a set of five modern-day OT models and propagated onto the level of observations. This error information is then employed within the least-squares parameter estimation for the final gravity product. The functionality of this approach is verified through closed-loop simulations and its limitations are discussed. Results are presented for single-pair-based gravity retrieval scenarios.