Estimating Earth's temporal gravity field from GRACE observations: Mitigation of thermal errors and the interplay between orbital characteristics, basis functions and spatial resolution

Abstract

The Gravity Recovery and Climate Experiment (GRACE) mission measured the combined effect of the Earth's static and time-variable gravity fields globally and near-continuously over 15 years at unprecedented accuracy. Launched in 2002, the GRACE mission used a unique low-Earth orbit satellite-to-satellite tracking mission design. The time-variable gravity field is influenced by the movement of masses within the hydrosphere and the solid Earth. By directly monitoring mass balance changes due to flood, drought, groundwater extraction, ocean circulation, ocean mass increase and ice mass loss, results from the GRACE mission have increased understanding of the impacts of human activity, natural variation and climate change. The accuracy of GRACE estimates of the time-variable gravity field and the associated mass anomaly time series is affected by several factors. These include orbital characteristics, quality of the observations and background forcing models and the processing strategies used for precise orbit determination and temporal gravity field estimation. This study aims to improve GRACE-based estimates of the time-variable gravity field to analyse mass anomalies by mitigating measurement errors and optimising the choices of processing strategies. The precise calculation of GRACE satellite orbits is reliant on knowledge of accurate non-gravitational forces acting on the spacecraft. Optimal performance of the accelerometers requires a highly stable thermally controlled environment which was not maintained throughout the mission. In this study, I developed pre-processing and calibration strategies to account for thermally-induced errors in the non-gravitational acceleration measurements. Accurate time-variable gravity models could then be estimated from GRACE data even in the presence of thermally-induced error. Some mathematical form, or basis function, must be assumed to parameterise the temporal gravity field on the surface of Earth. The choice of the inter-satellite observation and basis functions can also improve the recovery of the gravity field by better localising the mass variations. This study demonstrates how mass concentration (mascon) tiles can reduce signal leakage and intra-mascon variability (the variations of mass change signals within a mascon). I identified the optimal mascon parameterisation through simulation, subsequently used to generate the ANU GRACE mass anomaly time series. Improved localisation of the mass variation signals was achieved using the range acceleration as the inter-satellite observation rather than the conventional approach of using range rate observations. The GRACE processing strategies chosen to optimise the accuracy of the temporal gravity solutions tend to be used - without change - across the mission's duration. However, as the geometry of the orbits of the twin spacecraft vary throughout the mission, the ability of the observations to recover high-frequency spatial signals also varies. Through simulation, I assessed the impact of the changing orbital elements on the spatial resolution of the GRACE mascon solutions as a function of altitude and ground track density. With appropriate regularisation, mascons as small as $\sim$150 $\times$ 150 km yield the most accurate solutions even during periods of orbit resonance. Under realistic simulation conditions, the temporal gravity field solutions are significantly improved with decreased orbit altitude. Many components of my work have been implemented into the ANU GRACE software, including pre-processing and calibration strategies that account for thermally-induced errors in the accelerometer measurements, filtering to mitigate high-frequency inter-satellite range acceleration noise, protocols to create mascon grids and the iteration procedure used to generate the ANU GRACE mass anomaly estimates. The results show substantial seasonal variations, ice sheet mass loss and global mean sea level increase consistent with previous studies.