Reconstruction of GRACE Mass Change Time Series Using a Bayesian Framework

Area of Science:

• Geodesy and Earth Science
• Climate Science
• Data Science

Background:

• The Gravity Recovery and Climate Experiment (GRACE) and its Follow-On (GRACE-FO) missions revolutionized Earth's gravity field monitoring.
• Gaps in GRACE-FO data (33 monthly solutions) hinder continuous observation and analysis of temporal gravity changes.

Purpose of the Study:

• To impute missing GRACE-FO monthly solutions using Bayesian inference.
• To quantify uncertainties in both existing and newly generated GRACE-FO data.
• To provide a continuous, accurate dataset for Earth system science applications.

Main Methods:

• Modeled 196 GRACE-FO solutions (04/2002-04/2021) using an additive generative model.
• Employed Bayesian inference with informative priors and Markov Chain Monte Carlo (MCMC) for parameter estimation.
• Reconstructed 229 solutions by combining posterior medians and original residuals.

Main Results:

• Reconstructed GRACE-FO solutions captured 99% of basin-scale and 78% of one-degree grid-scale variability.
• The imputed data demonstrated superior accuracy compared to existing reconstructions in land surface modeling.
• The method provides a data-driven uncertainty quantification from the data generation process.

Conclusions:

• The Bayesian approach effectively reconstructs missing GRACE-FO data, ensuring data continuity.
• The enhanced dataset improves the accuracy of Earth system models and land surface studies.
• The predictive posterior distribution offers potential for near-real-time applications like data assimilation, reducing mission latency.