Abstract:

The high degree and order spherical harmonics of the time-variable gravity fields observed by the Gravity Recovery and Climate Experiment (GRACE) twin satellites gravity mission are dominated by noise. We construct two optimized variance-dependent smoothing methods that can more effectively reduce the high degree and order errors than the commonly used Gaussian smoothing. These optimized smoothing methods can maximize the signal-to-noise ratio between GRACE recovered terrestrial water storage change and residual signal and noise over the ocean. With the optimized variance-dependent smoothing, GRACE estimated global terrestrial water storages show significantly improved spatial resolution and less leakage error than similar results from the Gaussian smoothing and other published optimal smoothing techniques. Based on tests on advanced land surface model estimates, the equivalent spatial resolution from the optimized variance-dependent smoothing is about 500 km as compared to ~ 800 – 1000 km in the Gaussian smoothing.