Sediment yield estimation in ungauged catchments: a case of Tanzania

The study investigates the possibility of employing statistical techniques to derive a sediment yield predictive equation to be used to estimate long term sediment yield in ungauged catchments in Tanzania based on hydrologic and geormorphic parameters established from gauged catchments. At least 5 different numerical models are assumed and calibrated for each set of data to establish the existence of statistical relationships between sediment concentration and discharge, and rediment concentration and rainfall. The Linear Perturbation Model was used to simulate discharge data from daily rainfall data. From this daily discharge values were extracted and used in two of the numerical models having simulated discharge as one of the independent variables. The model results gave rather low R2 values, and Index of Volumetric Fitting (IVF) values generally close to unity. Based on the findings of this work, no strong statistical relationships exist between the dependent and independent variables of the models calibrated. This was demonstrated by the poor model calibration co-efficient of determination (R2), i.e. R2 less than 50%, for at least 9 out of 13 'best' models selected for the thirteen respective stations used, rendering the sediment data unusable. The predictive sediment yield equation envisaged could not be developed as the long term sediment yield for the majority of the catchments could not be ascertained from the rating equations with reasonable accuracy. The investigation leads to the conclusion that statistical equations developed in this form may not be applicable to the Tanzania case with the data available.