Water Storage in Africa from the Optimised GRACE Monthly Models: Iterative Approach

Abstract

Water storage variability in southern Africa and particularly in the Zambezi river basin is evaluated using optimally smoothed GRACE gravity field models recently developed at Delft University of Technology. Poor availability and low quality of hydrological in situ data make independent GRACE estimates valuable for hydrological modeling.The output of available hydrological models in the target areas is therefore used for the quantification of the sample correlation and the main discrepancy between the water storage estimates from GRACE and hydrology. Moreover, an attempt to identify the main sources of the discrepancy is made.The results of the study show the maximum sample correlation between optimal water storage estimates from GRACE and from the Lumped Elementary Watershed (LEW) regional hydrological model in the North and North-East of the Zambezi river basin. The maximum discrepancy of about 0.025 m between the mean water storage variations over the Zambezi river basin from LEW and GRACE has been observed in spring, when the water storage is the largest.The estimated signal leakage (bias) caused by the optimal filtering is practically negligible when compared with the GRACE estimates produced by other research centers, though it is considerable for hydrological applications and would require a bias correction for areas smaller then $0.5 \cdot 10^6 km^2$.A large discrepancy between LEW regional hydrological models of release 2008 (LEW-R2008) and 2006 (LEW-R2006) has been unexpectedly observed, especially in fall 2004 and spring 2005. This finding is presumably related to the use of the suspected higher quality of TRMM rainfall data with respect to FEWS rainfall data, respectively.It is finally concluded that the optimal GRACE estimates can be beneficially used to constrain regional hydrological models for their further improvement.