Hydrological Modelling of Ethiopian Watersheds : The Application of Hydrological Models for Flow Forecasting and Analysis of Sensitivity to Climate and Landuse Changes

Abstract

Hydrological modelling studies vary depending on the purpose, data availability and number of outputs required. Quantifying the hydrological responses to various human- and climate-induced changes requires different modelling schemes than those used for simple flow estimation. Based on these fundamental principles of model use, this research examines three major model applications. In part I of the research, a sensitivity analysis of Soil and Water Assessment Tool (SWAT)-simulated streamflow to hypothetical and Global Climate Model (GCM)-generated climate change scenarios was conducted within Eastern Nile basins (Blue Nile, Tekeze and Baro Akobo), which contribute to about 70% of Ethiopia`s total annual surface water potential. In part II, the effects of climatic and/or land use changes on the hydrology of the highly forested Omo Gibe river basin in southern Ethiopia were investigated. The study area is selected because one-third of the country’s power consumption is covered by this basin by means of constructed cascade hydroelectric power plants. In part III, simple and complex hydrological models were evaluated at a relatively smaller watershed (Gilgel Abbay) to revitalize systems-type black-box models for the purpose of flow forecasting. In Paper I, the hydrological model SWAT was run with daily and monthly precipitation and temperature data for the three basins of the Eastern Nile: the Abbay (Blue Nile), Baro Akobo and Tekeze basins. The model was calibrated and validated for the daily and monthly streamflow, as presented in the research paper by Mengistu and Sorteberg (2012). Twenty hypothetical climate change scenarios (perturbed temperatures and precipitation) were used to test the sensitivity of SWAT-simulated annual streamflow. The results reveal that the annual streamflow sensitivity to changes in precipitation and temperature differed among the basins and that the dependence of the response to the strength of the changes was not linear. On average, the annual streamflow responses to a change in precipitation with no temperature change were 19%, 17% and 26% per 10% change in precipitation, while the average annual streamflow responses to a 10% change in temperature and no precipitation change were −4.4% K−1, −6.4% K−1, and −1.3% K−1 for the Abbay, Baro Akobo and Tekeze river basins, respectively. In addition, 47 temperature and precipitation scenarios from 19 AOGCMs participating in CMIP3 were used to understand future changes in streamflow due to climate changes (Mengistu & Sorteberg, 2012). The climate models were in disagreement regarding both the strength and direction of future precipitation changes. Thus, no clear conclusions could be made about future changes in the Eastern Nile streamflow. However, such types of assessment are important as they emphasize the need to use several ensembles of AOGCMs, as the results are strongly dependent on the choice of climate models. In Paper II, the sensitivity of the Omo Gibe river basin in southern Ethiopia to both climatic and land use changes was investigated using the hydrological model SWAT. The model was calibrated and validated using observational data. Almost 60% of the average annual rainfall is lost through evaporation in the basin and the average runoff-rainfall coefficient was 0.26. Around two-thirds of the water yield was estimated to come from surface runoff, while groundwater was found to be responsible for the other third. The sensitivity of streamflow to precipitation changes was found to be high compared to the sensitivity to land use. On average, there was a 25% change in streamflow for a 10% change in precipitation. On the other hand, the response of streamflow to changes in temperature while holding the precipitation fixed is modest. A linear regression analysis of streamflow responses to the different temperature scenarios indicates that a 1°C change in temperature produces a 1.4% change in annual streamflow. The simulated effect of land use changes resulting from various hypothetical land use modifications was secondary to the effect of precipitation changes on the annual streamflow. However, the seasonal changes in streamflow were in some cases strongly affected by land use. For example, a deforestation scenario (entire forest-area coverage changed to bare lands) increased the January-April (dry season) streamflow by 38%. Results further indicate that the combined effects of land use and climate change may differ from the sum of the individual land use and climate change simulations. For example, in an increased precipitation scenario, changing land use to more bare land areas would increase streamflow and water yield less than simple additions of the individual effects. This shows that according to the model, nonlinear interactions among the water-balance components may occur when simultaneous changes in land use and climate change are imposed. From these two applications, streamflow proves highly sensitive to climate change scenarios and particularly to precipitation. For a unit change in precipitation, the change in streamflow is nearly double. Paper III evaluates the performance of different black-box rainfall-runoff models (simple single input-output models) and complex hydrological models for flow forecasting of the Gilgel Abbay catchment (upper Blue Nile river basin, Ethiopia). Seven black-box models embedded in the Galway River Flow Forecasting System (GFMFS) software packages, as well as HBV and SWAT models were applied. The performance of the simple linear model (SLM) is inferior to that of all other models. However, in comparison to the complex hydrological models (SMAR, HBV and SWAT), the simple single input-output models of the artificial neural network (ANN) and the linear perturbation model (LPM) outperformed in both simulation and updated mode when evaluated based on the statistical criteria of Nash-Sutcliffe Efficiency (NSE) and R2 at the Gilgel Abbay catchment. This indicates increasing model complexity (thereby increasing the number of tunable parameters), which does not necessarily enhance the model’s performance. This is clearly reflected in HBV and SWAT, with the performances accounting 86% and 66% their variances, respectively. Therefore, insofar as flow forecasting is concerned with smaller catchments, simple models are efficient and can be used for water development planning and management, thereby avoiding the difficulty of the calibration/validation of complex hydrological models. Furthermore, these types of models can be applied in data-scarce areas within Africa, thereby avoiding multi-parameter complex hydrological models.