Data-driven multi-step self-de-aliasing approach for GRACE and GRACE-FO data processing

Abstract

Temporal aliasing errors resulting from the undersampling of non-tidal atmospheric as well as oceanic mass variations constitute the largest limitation towards the retrieval of monthly gravity solutions based on GRACE and GRACE-FO satellite gravity missions. Their mitigation is thus a primary goal of current research. Unfortunately, the two-step co-parametrization approach proposed for application in Bender-type gravity retrieval scenario in Wiese et al. yields no added value for a single satellite pair. A detailed study of this parametrization strategy is carried out and it is shown that the reason for this is the flawed central assumption of the proposed method, that is that signals of different spatial wavelengths can be perfectly captured and separated with respect to their temporal extent. Based on this finding, we derive a multi-step self-de-aliasing approach (DMD) which aims to rectify the shortcoming of the Wiese et al. method specifically for the single-pair case while retaining its independence from background-model-based de-aliasing of non-tidal atmosphere and ocean (AO) signal components. The functionality and added value of this novel approach is validated within a set of numerical closed-loop simulations as well as in real GRACE and GRACE-FO data processing. The simulation results show that the DMD may improve the gravity retrieval performance in the high-degree spectrum by more than one order of magnitude if one aims to retrieve the full AOHIS (i.e. atmosphere, ocean, hydrology, ice, solid earth) signal, and by at least a factor 5 if a priori AO de-aliasing is applied. Simultaneously, the DMD is shown to degrade the retrieval of the low degrees, but it is also demonstrated that this issue can be mitigated by introducing a constraint into the processing scheme. The simulation results are widely confirmed by results obtained from applying the DMD to real GRACE/GRACE-FO data of the test years 2007, 2014 and 2019. The applicability of the DMD is further shown for Bender-type gravity retrieval. It is demonstrated that in case of a double-pair-based gravity retrieval this approach is at least equivalent to the Wiese et al. method.