Abstract
Leakage errors derived from spatial filters are the major limitation for estimating terrestrial water storage via the Gravity Recovery and Climate Experiment (GRACE) mission and the recently launched GRACE Follow-On mission. Here we develop an improved forward modeling method and assess its performance of reducing leakage errors over Africa. In noise-free condition, the forward modeling method shows its outperformance in restoring signals, and the improved forward modeling method can further reduce the leakage errors along the coastline of Africa. In noise-contaminated condition, the simulated environment is set as real as possible to GRACE mission and GRACE Follow-On mission. The results based on the simulated GRACE and GRACE Follow-On solutions demonstrate the capacity of improved forward modeling method in reducing leakage errors. In the case of simulated GRACE data, the average improvements of 24 basins over Africa are respectively 37% for annual amplitudes and 36% for trends. When compared with these simulated GRACE data, the improvements via simulated GRACE Follow-On solutions are minor over large and medium size river basins, but they are significant over small size river basins. In the case of simulated GRACE Follow-On solutions, the average improvements over Africa are 39% for annual amplitudes and 41% for trends. Eventually, the improved forward modeling method is used to process GRACE spherical harmonic datasets from the Center for Space Research (CSR). The results present better agreement with those derived from the newly released mascon solutions from Jet Propulsion Laboratory (JPL) and CSR, when compared with those derived from CSR Tellus grids with scale factors. The better consistency between these model-independent approaches indicates the good performance of our improved forward modeling method and the further necessity of careful evaluation of model-dependent approaches when using different prior hydrological models. Overall, experiments based on noise-free observations, noise-contaminated observations, and GRACE datasets indicate that improved forward modeling method is capable of restoring temporal signals.
Highlights
The Gravity Recovery and Climate Experiment (GRACE) mission has provided information about Earth’s time-variable gravity field with unprecedented accuracy between 2002 and 2017 (Tapley et al, 2004)
In noise-free condition, Chen et al (2015) have evaluated the performance of forward modeling method in West Antarctica, where the mass variations are only defined in two isolated regions
It is still necessary to evaluate the performance of the forward modeling method over Africa in noise-free condition
Summary
The Gravity Recovery and Climate Experiment (GRACE) mission has provided information about Earth’s time-variable gravity field with unprecedented accuracy between 2002 and 2017 (Tapley et al, 2004). GRACE data is contaminated by both observation errors and prior force model errors, including longitudinal stripes associated with correlations of certain SHCs separately as functions of even and odd degrees, aliases of tide model errors, and errors in oceanic and atmospheric model estimates To alleviate these errors, spatial smoothing filters as well as decorrelation filters are commonly used (Jekeli et al, 1981; Swenson and Wahr, 2006). General Forward Modeling Method Truncation and filtering of GRACE data are expected to significantly affect the amplitudes and trends of the true signals (Chen et al, 2015). To reduce these leakage errors, unconstrained forward modeling method is used, and the general procedures are as follows (Figure 2):. More details about the closed-loop simulation are presented in
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
