Detection and Modeling of the Effect of Large Earthquakes in GRACE/GRACE‐FO Mascons

Abstract The usefulness of Gravity Recovery and Climate Experiment (GRACE) and GRACE Follow‐On (GRACE/FO) data products for ocean studies remains limited in a few areas, where large gravity signals from the 2011 Japan earthquake and three major earthquakes in the Andaman‐Sumatra area (in 2004, 2007, and 2012) obscure vastly smaller signals of interest related to ocean dynamics. Furthermore, these earthquake effects can not be simply removed from the GRACE/FO products via a model during post‐processing, due to the loosening of constraints needed to allow the signal into current GRACE/FO products. Here, we create a multi‐earthquake empirical model from GRACE/GRACE‐Follow On data, estimating the equivalent water thickness changes as seen in the Center for Space Research (CSR) mascons due to the four major earthquakes mentioned. An iterative principal component analysis was used to compute piecewise linear approximations of the signals in these regions, which describe the co‐seismic bias changes during the known month of each earthquake as well as post‐seismic trends in local mass after/between quakes. This model is then applied as part of the estimation background field during a new iteration of CSR mascon processing designed for oceanic use, using tightened regularization to reduce noise in the near‐earthquake regions. By removing the CSR‐derived earthquake model and reducing month‐to‐month variability via the new regularization, we have reduced the GRACE/FO RMS in ocean mascons near the quake epicenters from more than 50 cm to under 5 cm, comparable to ocean mascon signals elsewhere. The resulting mascon series is more suitable for oceanic studies near Japan and Andaman Bay.