The Gravity Recovery and Climate Experiment (GRACE) mission has monitored total water storage anomalies (TWSA) globally with unprecedented resolution and accuracy since 2002. However, many applications require a data-based, multi-decadal extended record of TWSA prior to the GRACE period and for bridging the eleven-months gap between GRACE and its successor GRACE Follow-On (GRACE-FO), that does not depend on hydrological modelling. Statistical and machine-learning ‘reconstruction’ approaches have been developed to this end, mostly via identifying relations of GRACE-derived TWSA to climate variables, and some regional or global land data sets are now publicly available.
In this contribution, we compare the two global reconstructions by Humphrey and Gudmundsson (2019) and Li et al.(2021) mutually and against output from the water Global Analysis and Prognosis (WaterGAP) hydrological model from 1979 onwards, against large-scale mass-change derived from geodetic satellite laser ranging (SLR) from 1992 onwards, and finally against differing GRACE/-FO solutions from 2002 onwards.
We find that the reconstructions agree surprisingly well in many regions at seasonal and sub-seasonal timescales, even in the pre-GRACE era. We find larger differences at inter-annual timescales which we speculate are in part due to the way reconstructions are trained, and in part on which specific GRACE solution they are trained as well as the climatological characteristics of the region. Our comparison against independent SLR data reveals that reconstructions (only) partially succeed in representing anomalous TWSA for regions that are influenced by large climate modes such as El Ni$\tilde{\text{n}}$o-Southern Oscillation (ENSO).