Evaluation of High-Resolution Satellite Rainfall Estimates in Blue Nile River Basin, Ethiopia: A Hydrological Perspective

Accurate measurement of rainfall has paramount importance for the water-based economy of Ethiopia. However, the ground based rainfall measurement is limited to sparsely distributed rain gauges primarily located across the densely populated regions of the county. Due to their poor coverage these gauges hardly capture the spatially variable and topographically controlled rainfall of the region. The past two decades have opened a new chapter of global data acquisition from space using wide ranges of satellites including geostationary and polar orbiting satellites. The satellite-based rainfall estimate provides an opportunity to fill in the huge data voids of complex terrain regions in Ethiopia with less/no ground based measurements such as the Blue Nile River Basin. The satellite rainfall itself, on the other hand, is affected by uncertainties mainly resulting from large satellite sampling periods, indirect rainfall sensors and retrieval algorithms.;The objective of this study is to evaluate accuracy of four widely used high-resolution satellite rainfall estimates for hydrological applications in the Blue Nile River Basin. These satellite rainfall estimates are available at high spatial (< 25 km) and temporal (< 3 hr) resolution which are believed to be good for hydrological applications. The evaluations were conducted in two-pronged approaches. In the first approach, the evaluation was conducted through comparison of the satellite rainfall estimates with independent measurement. The independent observations were obtained through deployment of dense rain gauges in Ethiopia highland within scales equivalent to satellite rainfall pixels and using existing historical measurements obtained from sparse rain gauges in the proximity of our study watersheds. In the second approach, observed streamflow was used as a way of evaluating accuracy of satellite rainfall estimates in several watersheds through multiple hydrological modeling frameworks. This approach was also used to evaluate the propagation of satellite rainfall errors to simulated streamflow as a function of satellite rainfall algorithms, watershed size, and hydrologic model complexity levels.;The results showed that there are significant bias in satellite rainfall estimates, large variations among them and the simulation outputs. The major conclusions are as follows. i) Infrared-based and Satellite-Gauge merged satellite rainfall products severely underestimate rainfall in the mountainous regions of the Blue Nile; ii) microwave-based Satellite-Only algorithms provide reasonably accurate rainfall estimation in high elevation areas; iii) Satellite-Gauge merging algorithm provides relatively accurate estimate in low elevation areas where rain gauge density is low but unexpectedly severely amplifies errors in high elevation region; iv) rain gauge calibrated hydrologic model simulations amplify errors in satellite rainfall estimates; and v) calibrating the models using the corresponding satellite rainfall as inputs and using streamflow as the only objective function significantly improves the performance of simulated streamflow but decreased the performance of other water balance components. This study showed that accuracy of streamflow simulation is not always the reflection of accuracy of satellite rainfall estimates because of the effects of calibration on the propagation of satellite errors to other water balance components other than streamflow. Therefore, caution must be exercised when using streamflow as a way of evaluating satellite rainfall.