Comparing TRMM 3B42, CFSR and ground-based rainfall estimates as input for hydrological models, in data scarce regions: the Upper Blue Nile Basin, Ethiopia
A. W. Worqlul1,2,3, A. S. Collick4, S. A. Tilahun3, S. Langan2, T. H. M. Rientjes5, and T. S. Steenhuis1,31Department of Biological and Environmental Engineering, Cornell University, Ithaca, New York, USA 2International Water Management Institute, Addis Ababa, Ethiopia 3School of Civil and Water Resource Engineering, Bahir Dar University, Bahir Dar, Ethiopia 4USDA-ARS, University Park, Pennsylvania, USA 5Department of Water Resources, Faculty of Geo-information Science and Earth Observation (ITC), University of Twente, Enschede, the Netherlands
Received: 27 Dec 2014 – Accepted for review: 01 Feb 2015 – Discussion started: 18 Feb 2015
Abstract. Accurate prediction of hydrological models requires accurate spatial and temporal distribution of rainfall observation network. In developing countries rainfall observation station network are sparse and unevenly distributed. Satellite-based products have the potential to overcome these shortcomings. The objective of this study is to compare the advantages and the limitation of commonly used high-resolution satellite rainfall products as input to hydrological models as compared to sparsely populated network of rain gauges. For this comparison we use two semi-distributed hydrological models Hydrologiska Byråns Vattenbalansavdelning (HBV) and Parameter Efficient Distributed (PED) that performed well in Ethiopian highlands in two watersheds: the Gilgel Abay with relatively dense network and Main Beles with relatively scarce rain gauge stations. Both are located in the Upper Blue Nile Basin. The two models are calibrated with the observed discharge from 1994 to 2003 and validated from 2004 to 2006. Satellite rainfall estimates used includes Climate Forecast System Reanalysis (CFSR), Tropical Rainfall Measuring Mission (TRMM) 3B42 version 7 and ground rainfall measurements. The results indicated that both the gauged and the CFSR precipitation estimates were able to reproduce the stream flow well for both models and both watershed. TRMM 3B42 performed poorly with Nash Sutcliffe values less than 0.1. As expected the HBV model performed slightly better than the PED model, because HBV divides the watershed into sub-basins resulting in a greater number of calibration parameters. The simulated discharge for the Gilgel Abay was better than for the less well endowed (rain gauge wise) Main Beles. Finally surprisingly, the ground based gauge performed better for both watersheds (with the exception of extreme events) than TRMM and CFSR satellite rainfall estimates. Undoubtedly in the future, when improved satellite products will become available, this will change.
Worqlul, A. W., Collick, A. S., Tilahun, S. A., Langan, S., Rientjes, T. H. M., and Steenhuis, T. S.: Comparing TRMM 3B42, CFSR and ground-based rainfall estimates as input for hydrological models, in data scarce regions: the Upper Blue Nile Basin, Ethiopia, Hydrol. Earth Syst. Sci. Discuss., 12, 2081-2112, doi:10.5194/hessd-12-2081-2015, 2015.