Model for predicting sedimentation of reservoirs in Semi-Arid areas of Uganda

Abstract

Prediction of reservoir sediment yield and identification of the most significant factors affecting the rate of reservoir sedimentation continue to be the main challenges faced by government agencies and water utility companies. This is mainly due to lack of data and suitable models. The use of physical and conceptual models is constrained by lack of data for model calibration and validation. Whereas many sediment yield prediction models have been developed, these have not been validated in ungauged catchments due to lack of data. Moreover, these studies have been conducted elsewhere with different physiographic properties and climatic conditions. Therefore, there is need to further improve prediction of sedimentation in reservoirs. This study was aimed at developing a multiple regression model for predicting the sedimentation rates of reservoirs in the semi-arid areas in Uganda through assessment and characterisation of reservoir catchments, estimating the sediment yield and there after developing the model. The reservoir catchments were delineated using STRM DEM in GIS environment. The independent variables of the model were reservoir Catchment Area (CA), Slope (SL), Rainfall (RF) in mm, Soil Condition (SC), Drainage Network (DN) and Land-use(LU). The qualitative variables (SC, DN & LU) were characterised by scoring using a computer programme coded in C # programming language in visual studio. The findings of the study showed that the reservoirs in the study area experience severe sedimentation, with sediment yield values ranging between 37.1 to 77.5 tons/km2/year with an average of 61.7 tons/km2/year. These estimated sediment yields were compared with those estimated by the commonly used RUSLE method (mean 66.4t/km2/y). There were no significant differences between the values of sediment yield computed by the model and those computed by RUSLE model thus validating the developed model. The study developed the regression model; SY=1.663SL+6.267LU−1.841DN−0.29SC+100.112CA+0.226RF−491.634 (R2=0.962; n=17; CI of 61.7±0.1146 tons/km2/year). The model was found to predict sediment yield accurately in the study area with a model efficiency (NSE) and coefficient of determination (R2) of 0.893 and 0.962 respectively. The model could be helpful for practitioners to quickly predict sediment in ungauged catchments in the semi-arid areas of Uganda.