Sediment Yield Prediction in Ungauged Catchment and its Impact on Reservoir Development in Ethiopia Highlands
No Thumbnail Available
Date
2026-03-01
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Predicting sediment load for ungauged catchments remains the most uncertain and complex problem in hydrology. A clear understanding of sediment yield, transport, and deposition processes
in river basins is critical to developing efficient measures for mitigating reservoir sedimentation and predicting reservoir lifespans. Sedimentation, a major issue, significantly affects the
sustainability and operational efficiency of reservoirs and deteriorates dam operation. Reservoir sedimentation is heavily influenced by the supply of sediments from catchments in the upper
reaches, influencing riverbank deposition of sediments as well.
Although reliable sediment data is vital for the planning of effective dams and water resource systems, most parts of the world, especially in developing countries like Ethiopia, lack
hydrological monitoring and sediment data. This is especially true in the highlands of the Awash and Abbay (Blue Nile) river basins, where sediment measurements are limited. Meanwhile, these basins are home to important water-related projects such as the Grand Ethiopian Renaissance Dam (GERD) and several proposed irrigation and hydropower initiatives. Given this gap, there is a pressing need to develop viable alternative models for estimating suspended sediment output. The main objective of this study was to estimate suspended sediment yield in the ungauged Abbay and Awash catchments. To achieve this, the research specifically aimed to (1) assess the effects of land use and land cover changes (LULC) on sediment yield dynamics in the Upper Awash River Basin using the QGIS-based Soil and Water Assessment Tool PLUS model (QSWATPLUS), (2) analyze the spatial and temporal variability of suspended sediment yield in the Upper Blue Nile Basin using QSWATPLUS, (3) develop an alternative empirical model for estimating suspended sediment yield using conceptual model, and (4) compare different modeling approaches used for suspended sediment yield prediction.
In this study, the Quantum Geography Information System Interference Soil and Water Assessment Tool Plus (QSWAT-PLUS) model was used to assess the impacts of land use and land
cover (LULC) on sediment load in the Upper Awash River Basin (UARB), which is experiencing sedimentation issues in the Koka reservoir. The Modified Universal Soil Loss Equation within the
SWAT Plus model was employed to simulate streamflow and sediment yield, thereby identifying spatiotemporal hotspots of sediment variability across various reservoir catchments. The QSWAT Plus model incorporated digital elevation models, climate data, land use, soil types, and slope characteristics of the Upper Blue Nile Basin (UBNB). The calibration and validation of monthly streamflow and sediment yield were conducted using the Sobol tool algorithm from the SWAT Toolbox. Furthermore, the QSWAT+ model divided the Kessie watershed into 18 sub-basins in order to analyze each catchment's characteristics. The performance testing of the alternative model utilized a data set spanning 11 years from six watersheds, the assessment of which was conducted through the utilization of model statistics. Principal component analysis (PCA) was used to identify significant variables affecting sediment yield. In addition, data reduction techniques, including the Gamma test (GT), classification and regression trees (CART), and stepwise regression (SR), were used to select the most influential factors on suspended sediment output. Finally, multiple linear regression (MLR) and artificial neural networks (ANN) were used to develop a regional model for estimating suspended sediment yield (SSY) in ungauged catchments.
Results indicated that the mean annual sediment yield entering the Koka reservoir under the 2005, 2010, and 2015 LULC scenarios was about 26.03, 26.34, and 28.33 t/ha/yr, respectively. In
general, streamflow, surface runoff, and sediment output increased by 4.5%, 12.68%, and 8.84%, respectively, due to the rapid change of LULC from 2000 to 2015. Temporally, the sediment yield at the upstream side of the Koka Dam watershed was 60.8% during the wet season. The upward trend indicates that changes in LULC bring increased sediment loads, which may endanger reservoir sustainability and watershed health. Simulation outputs revealed that filter strips, contour farming, and terracing best management practices (BMPs) decreased sediment yield by 60%, 65%, and 80%, respectively. Terraces were most effective in controlling erosion in the priority Sub basin upstream of the Koka Dam.
Spatial variability in sediment yield within the Kessie Sub-basin was observed to range extensively from 0 to 67.6 t/ha/yr. A further analysis revealed that 42.04% of the watershed falls within the critical erosion zone and 39.48% falls within the sub-critical zone. This spatial variability is attributed to a combination of environmental factors and human activities. It is imperative to note
that this critical amount of sediment yield has the potential to impact the service life of the reservoir. It is estimated that approximately 90% of the annual sediment load was transported
during the wet season. These BMPs can be effectively used to minimize sediment transport, which controls reservoir sedimentation in the Upper Blue Nile Basin. Out of 24 variables screened, seven were found to be the most dominant significant factors: drainage area, stream slope, main channel length, rainfall, agricultural land cover, forest cover, and streamflow. Collectively, climatic, geomorphological, and hydrological variables were found to be major drivers of suspended sediment yield (SSY). Following calibration and validation of
streamflow and suspended sediment yield data at the Kessie gauge station using the QSWATPLUS model, a new empirical model was developed to estimate suspended sediment yield in the
ungauged catchment. The new empirical model developed was verified based on three statistical performance metrics, proving to be highly accurate and versatile in varying environmental
conditions, even in areas where data are limited. This improves its credibility in sediment management and watershed conservation planning.
In all the tested models, the Genetic Algorithm-based Artificial Neural Network (GT-ANN) achieved superior performance compared to the ANN and Multiple Linear Regression (MLR)
models. GT-ANN model under calibration (R² = 99.9%, RMSE = 12,631.6 t/yr, and MAE = 10,354.9 t/yr) and under validation (R² = 82%, RMSE = 139,944 t/yr, and MAE = 136,036 t/yr)
demonstrated that this model is effective in estimating sediment yield. Overall, determining influential SSY drivers and using AI-driven modeling can empower water resource managers to effectively estimate SSY amounts in ungauged basins, promoting improved sediment management and conservation plans.
Description
Keywords
Sediment Yield, Ungauged Catchment, Empirical Model, Suspended sediment yield, Physical Model, Artificial Niral Network, Data Reduction Technique