Description:
<jats:title>Abstract</jats:title><jats:p>The Gravity Recovery and Climate Experiment (GRACE) mission has monitored global total water storage anomalies (TWSA) with an unprecedented accuracy since 2002. Yet, many applications require a longer record, that is, extending prior to the GRACE period. Here, we present a new global reconstruction of long‐term (1979–2020) TWSA fields by combining machine learning with time series decomposition and statistical decomposition techniques. We find that the long‐term TWSA reconstructed from GRACE fits well with the GRACE‐FO observation over most grids (0.5° resolution) of the global land and successfully reproduces the strong El Niño signal. Comparisons to Satellite Laser Ranging solutions and to observed global mean sea level change suggest our reconstruction (doi: <jats:ext-link xmlns:xlink="http://www.w3.org/1999/xlink" xlink:href="https://doi.org/10.5061/dryad.z612jm6bt">https://doi.org/10.5061/dryad.z612jm6bt</jats:ext-link>) is more reliable than previously published products. This study provides a viable approach for both reconstructing past TWSA and filling the GRACE data gap at the global scale.</jats:p>