Data-driven Uncertainty-aware 5-day Surface Mass Variations from GRACE Ranging Residuals 

Abstract

This study directly utilises GRACE Level-1B (L1B) KBR post-fit residuals, provided by the Center for Space Research, as the basis for high-frequency mass change modelling. We showcase the updated methodology and its uncertainty characteristics, demonstrating the potential of KBR post-fit residuals to recover sub-monthly surface mass variations from the GRACE(-FO) mission by exploiting information already contained in the conventional monthly gravity-field processing chain, without requiring additional observations or a separate high-frequency gravity-field inversion. We submitted the work to the Journal of Geodesy, and it continues the methodology presented in our previous Science Meeting contribution. The methodology combines geo-fits with post-fit residuals from the KBR instrument of the GRACE satellites. Geo-fits represent the range-rate L1B observations fitted to monthly spherical harmonic gravity models in traditional least-squares gravity retrieval. The post-fit residuals represent the instantaneous discrepancies between the reduced observations (pre-fits) and the corresponding geo-fits, and therefore contain information not represented by the monthly gravity-field solution. We spatially interpolate 5-day sets of post-fit residuals and combine them with the corresponding spatially binned monthly geo-fits, resulting in gridded high-frequency (HF) geo-fits. We then transform the HF geo-fits into equivalent water height (EWH) using an empirical relationship. Since the previous study, the methodology has been further developed in several aspects. The updated framework considers residuals in both the range and line-of-sight gravity difference (LGD) domains, providing three formulations: Range, LGD, and a hybrid formulation. The latter uses a weighted and normalised combination of both observables, with the weighting selected to minimise the regression error in the transformation to EWH. The observation-to-EWH transformation has also been improved, showing that the relationship is not only spatially variable but also time-dependent, reflecting changes in inter-satellite range and satellite altitude throughout the GRACE mission. These variations affect KBR sensitivity to mass change but should not appear as EWH variations; the updated transformation therefore accounts for these effects. The three formulations (range, LGD and hybrid) are available as 5-day EWH products (https://doi.org/10.4121/4fc748e8-01c7-4f06-87da-653937b078f7). Further developments include uncertainty propagation, which provides spatially and temporally resolved uncertainty estimates for the 5-day EWH fields. The additional uncertainty associated with reducing the temporal resolution from monthly to 5-day estimates remains moderate. Relative to the corresponding monthly solutions, the uncertainty increases by approximately 13–20% over land and 8–12% over the oceans.  For comparison, reducing the averaging period from one month to five days while assuming independent observations and no additional information would yield an expected uncertainty increase of approximately 145%. The relatively small increase results from exploiting the spatial and temporal information contained in the KBR post-fit residuals rather than treating each 5-day solution as an independent gravity-field inversion. The approach does not use hydrological models or other external geophysical priors, nor does it require temporal regularisation or Kalman filtering. It therefore avoids the need for a separate high-frequency gravity inversion framework. The resulting 5-day EWH grids are evaluated against daily solutions and through case studies of the 2007 Bangladesh floods and the January 2013 Cyclone Oswald.