Long-term trend analysis of the hydrology and sediment load in the Upper Blue Nile basin, using SWAT model

The upper Blue Nile basin is crucial for the socio-economic development and environmental stability in the three riparian countries, Ethiopia, Sudan and Egypt. However, the intensified land use and land cover changes in the Ethiopian high lands have increased the inter annual and inter decadal variability of hydrological and sediment flows of the basin. The aim of this study was to understand the long-term variations of hydrology and sediment fluxes using statistical models (Man-Kendall and Pettitt tests) and to physically verify the results with the SWAT model. The gradual trend test (Man-Kendall), and abrupt change test (Pettitt) have been employed on seasonal and annual discharge of 1970-2009 and sediment load data series of 1980-2009 at the out-let of the basin near Ethiopia-Sudan border. Both tests showed that there has been an increasing trend of annual and wet season flow as well as sediment load, where as a decreasing trend of stream flow has occurred during the dry season. This excludes rainfall as a key driver for change of runoff and sediment fluxes. Next, a physically based model (SWAT) was developed and ran at the beginning and end of the time series span in order to verify causes of flux changes. The model was calibrated based on daily data for the period of 1970-1973. Optimized parameters from calibration period were validated in the period of 1974-76 and 2000-2005, respectively. The performance of the model was evaluated using Nash-Sutcliff efficiency (NSE), Coefficient of determination (R2) and percent bias (PBIAS). The computed NSE, R2, and PBIAS for calibration period were 0.82, 0.84, and -5.24% and validation (1974-1976) of daily flow found to be 0.8, 0.76, and 18.4%, respectively. Model performance indices for sediment load was found to be NSE=0.76, R2=0.78, and PBIAS=12.38% for calibration and NSE=0.73, R2=0.75 and 20.85% for Validation (1974-1976) period. This shows acceptable performance of SWAT model. The daily flow and sediment load showed unacceptable model performance in the 2000-2005 simulation periods. The model performance indices found to be NSE=0.18, R2=0.37 and PBIAS=28.3% for daily stream flow and NSE=0.46, R2=0.57 and PBIAS=28.3% for daily sediment load simulation. This indicates that optimum model parameters from 1970's are incapable of predicting stream flow and sediment load accurately unless the model is calibrated again for the 2000's simulation periods. This implies that there is model parameter variation between 1970's and 2000's simulation periods. The model captured the timing of all hydrographs and base flows nicely which clearly indicates that parameters in relation to river routing, sub surface and ground water flow have not changed for the last 39 years. This may lead to the conclusion that the change of model parameters are particularly the surface runoff response parameters which control peak flows and rate of soil erosion/ sediment transport. Accordingly, this study concludes that the observed change of stream flow and soil erosion/ sediment transport is the outcome of land use changes and degradation of the upper Blue Nile catchment