A Framework for the Assimilation of GRACE/GRACE-FO Observations into JPL Terrestrial Reference Frame Products

Abstract

Space gravimetry from GRACE and GRACE Follow-On (GRACE/GRACE-FO; GGFO)provides quasi-monthly estimates of global mass redistribution that canbe mapped into elastic surface displacements. Although spatially coarserthan the station-based observations underpinning terrestrial reference framesolutions, GGFO-derived displacements show substantial consistency with GNSSestimates, particularly in the vertical component, where correlations withIGS Repro3 solutions approach 0.7 for both CSR and JPL mascon products. Thiscomplementarity offers an opportunity to improve the representation ofnon-tidal surface deformation in terrestrial reference frame (TRF) solutions.JPL TRF products are generated using a square-root information Kalman filterthat assimilates product-level inputs from GNSS, VLBI, SLR, and DORIS withina common state-space framework. In JTRF2020 and subsequent realizations,non-tidal loading is represented primarily through co-estimated seasonal andother temporal components. Here we investigate an alternative formulation inwhich GGFO-inferred deformation is introduced as an exogenous control inputdriving the load-responsive components of the TRF state, so that gravimetrycontributes to state evolution rather than entering as an additionalobservation equation.Starting from Level-3 mascon solutions, surface mass fields are transformedinto the spherical-harmonic domain and adjusted to obtain a consistentrepresentation of time-variable gravity, including treatment of degree-1and other low-degree coefficients, model-dependent corrections, and removedand restored non-tidal geophysical variability. The framework can alsooperate directly on conventional spherical-harmonic gravity solutions. Gapsin monthly geopotential-coefficient time series are reconstructed usingmultichannel singular spectrum analysis (MSSA), followed by interpolationto daily cadence using piecewise cubic Hermite interpolation. The resultingdaily fields are converted to elastic surface displacements using load Lovenumbers and mapped to sites represented in the JPL frame solution.This approach also separates spatial scales within the TRF estimationproblem. GGFO constrains higher-degree load-induced deformation, whilesatellite laser ranging remains the principal technique for determiningdegree-1 geocenter motion. Incorporating GGFO-derived deformation into thestation-position state may therefore reduce aliasing between regional orcontinental loading and SLR-derived common-mode geocenter motion.We present the methodology, validation, and current status of theseassimilation efforts toward a unified state-space representation of terrestrialreference frame dynamics that combines space gravimetry and geometric geodesy.