Stream Flow Modeling and Uncertainty Analysis in Didessa River Basin, Blue Nile, Ethiopia

Abstract

To effectively plan for water resources and protect against watershed problems, it is necessary to understand the quantity and quality in space and time through studies. The general objective of this study is to model stream flow of Didessa River using SWAT model and to analyze the related uncertainties. Digital elevation model, Land use classification map, Soil map and the available weather data of 1980 to 2016 were used and the whole Didessa basin was separated into 674 hydrological response units (HRU) in 112 sub-watersheds. The available flow data of 1997-2014 was used for calibration and validation at 2 hydro gauging stations. SUFI-2 and GLUE program of SWAT CUP was used for calibration and uncertainty analysis and performance of the two programs in calibrating SWAT model was also compared. The SWAT model developed for the river basin evaluated and its performance is certain with the statistical measures, coefficient of determination (R2) and Nash and Sutcliffe coefficient (NS).The model performance was very good for monthly time steps. The obtained statistical results of (R2,NS) by SUFI-2 at Dembi were (0.75 ,0.74) and (0.78,0.78) for calibration and validation respectively. The obtained result at Arjo were (0.72,0.71) and (0.73, 0.72) for calibration and validation respectively. The obtained result by GLUE were (0.77 ,0.75) and (0.78,0.77) for calibration and validation at Dembi and (0.73,0.71) and (0.77, 0.72) for calibration and validation at Arjo. The performance of the model for daily time steps were also evaluated. The obtained result of (R2, NS) value for calibration and validation (0.72,0.69) ;(0.63,0.62) and (0.68,0.66) ;(0.62,0.62) for SUFI-2 and GLUE respectively at Dembi station. The result of uncertainty analysis done by SUFI-2 shows 40-48% and 24-44% percent of observed flow is bracketed by 95PPU for monthly and daily time steps respectively. GLUE uncertainty analysis program brackets 25-34% and 28-29% of observed flow for monthly and daily time steps respectively. GLUE uncertainty program able to obtain high value of R2 and NS with small percent of p and r-factor which shows good parameter identification, this shows that the overall associated uncertainty come from either conceptual or input or a combination of them but not from parameter identification. So, Both SUFI-2 and GLUE performs well in calibrating SWAT model and they were balanced predictive approach. The calibrated model can thus use for futuristic prediction and as an input for decision making in developing a better sufficient and integrated water resource management of the river basin