Evaluation of runoff estimation from GRACE coupled with different meteorological gridded products over the Upper Blue Nile Basin

Water balance closure using purely remote sensing products was difficult to achieve until the launch of Gravity Recovery and Climate Experiment (GRACE) satellites in 2002. The accurate quantification of water cycle components (precipitation, evapotranspiration, runoff, and terrestrial water storage) over a large-scale basin is an important step in improving the understanding of the water balance and the response of the basin to different hydrologic extremes. The Upper Blue Nile (UBN) basin contributes about 60% of the streamflow to the main Nile River annually, and hundreds of millions of people heavily rely on the Nile River. Thus, accurate quantification of the hydrological cycle fluxes will help manage the water resources in an effective, sustainable manner. Hydrometeorological data is lacking; nevertheless, remote sensing data provides an alternative approach to estimating the water cycle components. However, prior to incorporating these products into the water budget calculation, their performance over the studied basin should be assessed. In this study, we aim to estimate runoff from the water budget equation and diagnose the estimated runoff with the Eldiem gauge records at the outlet of the UBN basin for the 2003–2014 period. We evaluate the water cycle components for seven rainfall products (CHIRPSv2, CRU TS4.06, ERA5, TRMM 3B43 V7, GPM, CFSR, and SM2RAIN-CCI), three evapotranspiration products (GLEAM, MOD16, and PLM), and two terrestrial water storage solutions (GRACE JPL MASCON, and Spherical Harmonic (SH) products). The Overall Unified Metric (OUM) approach is adopted to choose the best performing combination among the 42 combination scenarios. The OUM is an approach based on summing up the rankings given for the error and linear fit metrics—namely, R², slope, y-intercept, RMSE, MAE, and PBIAS. Among the 42 combinations, the best rainfall, TWS, and ET combination performance products to estimate runoff are SM2RAIN-CCI, GLEAM, and GRACE SH, respectively. The statistical results for the six chosen metrics are R² = 0.7, slope = 1.6, y-intercept = - 0.5 cm, RMSE = 3 cm, MAE = 2.8 cm, and PBIAS = 36%. The 95% confidence bound of the combination scenarios was found to be able to bracket the runoff during the dry season, but the runoff was overestimated during the rainy season. The uncertainty analysis revealed that all the combinations were able to estimate the seasonal trend variation, but closing the water balance equation was not achieved. This deviation in closing the water budget equation might be attributed to the uncertainty associated with satellites, the limitation of land surface models to account for anthropogenic activities, and the coarse resolution of GRACE. Additionally, the signal processing uncertainties and the different algorithm assumptions of the remote sensing products may also have an influence. Further studies are needed to improve the reliability of the remote sensing product for the water budget closure, especially for applications on ungauged basins. Moreover, advancement in satellites will lead to accurate estimates in the near future.