Hydrological modelling of fine sediments in the Odzi River, Zimbabwe

Siltation of reservoirs is a major concern in Zimbabwe. Therefore, developm ent of prediction tools is of great importance. In the present study a rece ntly developed empirical sediment model (MV-SED) based on a daily rainfall- runoff model was applied to stimulate riverine fine sediment transport in a 2 486 km(2) catchment in eastern Zimbabwe. The model performance was evalu ated and changes in the model structure were suggested. The modelling was. however, associated with many uncertainties due to the adopted simplificati on of transport processes, An analysis of the model structure and a compari son with the rating curve function was done. The required length of data fo r calibration purposes was evaluated and model validation through split sam ple and proxy basin comparison was performed, Furthermore, since the empiri cal model was dependent on monitored runoff and fine sediment.,concentratio ns for calibration purposes, a field measurement campaign was conducted to assess the accuracy of observed data at the station studied,The field measu rements showed large errors in monitored runoff and fine sediment concentra tions for the 1998/99 wet season, which illustrated the uncertainty in pred ictions of fine sediment transport based on observed data. The HBV-SED mode l, which was applied over a period when data were believed to be fairly acc urate, simulated the fine sediment transport volume well for the validation period if it was calibrated for a minimum of four years. A shorter calibra tion period led to a significant increase in prediction uncertainty. The mo del failed to simulate individual high fine sediment peaks accurately mainl y due to poor performance of the rainfall-runoff model on a daily time-scal e even if the seasonal flow dynamics were described properly. In the studie d catchment the HBV-SED model application resulted in equally poor R-2-valu es as the rating curve technique, while thee estimated fine sediment volume was more accurate.