EVALUATION OF SATELLITE RAINFALL PRODUCTS FOR HYDROLOGICAL MODELING (THE CASE OF GREAT GIBE WATERSHED)

Abstract

Satellite-rainfall products are recognized as an essential source of rainfall data, especially in the region where ground based measurements are unavailable. Therefore, the main objective of this study is to evaluate the capabilities, applicability and limitations of satellite rainfall products such as CMORPH, TRMM 3B42v7 and PERSIANN, and inputs of hydrological models, including HBV-96 and SWAT-2012. These products and inputs were employed to simulate stream flow in the Great Gibe Watershed. The study period 2000-2012 was used for downloading and extracting the selected satellite rainfall estimates with daily-temporal and 0.250 x 0.250 spatial resolution. Sensitivity and uncertainty analysis, calibration and validation of the model were done using SWAT- CUP particularly the Sequential Uncertainty Fitting (SUFI-2) algorism for all rainfall inputs independently. The calibration period was from 2001-2009 leaving two year as a ―warm up‖ period and the validation period was from 2009-2012 for satellite rainfall based simulations as well as in situ based simulations. Based on the modeling results of SWAT and HBV models had showed better performance when calibrated with the in situ rainfall with model performance efficiency of R2 (0.9, 0.83), ENS (0.89, 0.81) and PBIAS (-8.2%, -8.3%) for SWAT and HBV models respectively while, the satellite rainfall estimates (TRMM-3B42v7 and CMORPH) showed relatively good performance when calibrated both models with R2 (0.77, 0.75), ENS (0.65, 0.68) and PBIAS (-23.1%, - 19.8%) for TRMM-3B42 and R2 (0.73, 0.65), ENS (0.56, 0.55) and PBIAS (-39.81%, - 28.6%) for CMORPH in both models respectively. But, PERSIANN satellite rainfall estimates showed poor performance for all models R2 (0.49, 0.48), ENS (0.51, 0.50) and PBIAS (-46.1%, -39.5%) before bias corrected. Although, satellite rainfall based simulations specifically TRMM3B42v7 based simulations capture the shape of observed stream flow hydrograph, there was under and overestimation of the stream flow volume simulated by satellite rainfall products followed by the reduction of model performance statistics. Bias corrected satellite rainfall based simulations had significantly improved the model performance as well as the volume of stream flow simulated {R2 (0.82, 0.75 and 0.59), ENS (0.73, 0.64 and 0.64) and PBIAS (-17.43%, -25.1% and -35.64)} for SWAT in case of TRMM-.3B42V7, CMORPH and PERSIANN respectively. For HBV model {R2

(0.81, 0.68 and 0.54), ENS (0.74, 0.6 and 0.53) and PBIAS (-15.3%, -24.9% and -31.35)} in case of TRMM-.3B42V7, CMORPH and PERSIANN respectively.