GRACE-like gridded reconstructions of total water storage change from SLR and statistical techniques

The total water storage change (TWSC) data record from the GRACE mission has found numerous applications in hydrology, oceanography, cryosphere, solid Earth and sea level research, and climate sciences. However, GRACE was only launched in 2002 and it lasted until 2017 with an increasing number of missing data, while GRACE-FO continued in May 2018. There will be no TWSC data within a gap of at least 11 months, and no GRACE-like data exists prior to 2002. As a result the current data record is not sufficiently long to for validating climate model predictions even for the long-term mean of TWSC, let alone for the probability of extreme events.

In this contribution we will discuss ways to reconstruct terrestrial TWSC from either geodetic or hydrometeorological data, or possibly from a combination. We will present new reconstructions of TWSC from 1992 onwards from satellite laser ranging (SLR) and from statistical learning methods, discuss various approaches, and compare to conventional SLR solutions, the Humphrey (2019) statistical reconstruction, and to GRACE in the more recent time frame. We show that both geodetic SLR analysis and/or multilinear regression or machine learning approaches can be successfully applied in a regularized framework where spatial modes from a GRACE-era TWSC decomposition inform the reconstructions. We show that these reconstructions reproduce many phenomena seen in modelling studies beyond the seasonal cycle; e.g. the suspected gradual onset of Greenland mass loss around 1998-2000, increase ocean mass rate since about 2011, or the presence of ENSO events.