Hydrological Models Comparison in Simulating Streamflow in the Melka Kunture Watershed in Upper Awash Sub-basin, Ethiopia

Abstract

Analysing hydro-meteorological variability and evaluating the performance of hydrological models are closely related processes in the field of hydrology. Understanding the variability of hydro-meteorological factors is crucial for accurate modeling and prediction of streamflow to overcome water-related challenges and enhance water resources management. This study aimed to identify a suitable hydrological model for simulating streamflow of Melka Kunture watershed in the Upper Awash sub-basin, in Ethiopia. The variability of historical rainfall and streamflow of the watershed was analysed using coefficient of variation, standard rainfall anomaly index, and precipitation concentration index. Three hydrological models (HEC-HMS, SWAT+, and HBV-Light) were considered, and their performances in simulating historical streamflow were evaluated using statistical model performance evaluation indices namely coefficient of determination (R2), Nash-Sutcliffe efficiency (NSE), and percent bias (PBIAS). Sixteen years (2000-2015) daily rainfall and streamflow data were used for models calibration and validation at the Melka Kunture watershed outlet. The HEC-HMS model was calibrated by the optimization trial (automatic calibration); the SWAT+ Toolbox was applied to analyse sensitivity analysis (Sobol method) and calibration (Dynamically Dimensioned search method); and in the HBV-Light model, Genetic Algorithm and Powel Optimization were used to calibrate and validate the model. Based on the results obtained, the variation of rainfall within the watershed ranges from low to high seasonally and annually. The result of the annual standard rainfall anomaly index showed the precipitation within the watershed ranges from extremely wet to extremely dry, but mostly all stations get near normal rainfall annually. The annual precipitation concentration index result within watershed ranges from irregular to moderate rainfall distribution, while seasonally, the precipitation concentration index within the watershed showed strong irregularities of rainfall. The performance of HEC-HMS, SWAT+, and HBV-Light in simulating historical streamflow of the watershed were found to be (R2: 0.71, NSE: 0.677, PBIAS: -0.1), (R2: 0.71, NSE: 0.669, PBIAS: 9.349), and (R2: 0.78, NSE: 0.8, PBIAS: 0.94), respectively during calibration. During validation, the values (R2: 0.68, NSE: 0.676, PBIAS: 6.39) for HEC-HMS, (R2: 0.68, NSE: 0.66, PBIAS: -0.475) for SWAT+, and (R2: 0.68, NSE: 0.68, PBIAS: 0.99) for HBV-Light were obtained. Based on the statistical indices, the HBV-Light model showed good correlation between observed and simulated daily streamflow during calibration and validation periods. It outperformed the other models during calibration, while the HEC-HMS model showed better performance in simulating low streamflow conditions during both calibration and validation periods, as indicated by the Nash Sutcliffe efficiency (NSElog) values of (HEC-HMS:0.70, SWAT+:0.19, and HBV-Light: 0.64; and HEC-HMS: 0.71, SWAT+: -2.07, and HBV-Light: 0.51) during calibration and validation, respectively. In conclusion, this study suggests that the HBV-Light model is the most suitable for simulating streamflow and HEC-HMS in simulating low streamflow conditions in the Melka Kunture watershed. Generally, the output of this study can be applied in planning, designing, and sustainable management of water resources in the Melka Kunture watershed.