Precipitation forecasts for hydrologic rainfall runoff prediction (case study in Ribb and Gumara rivers, Upper Blue Nile, Ethiopia)

Flow prediction is an essential part of flood early warning system which is one of the non structural measures against flooding. The reliability of the flow prediction process and the lead time it extends to mainly depend on the accuracy and availability of precipitation forecasts. At present this can be achieved with the help of Quantitative Precipitation Forecasts (QPFs) that can simulate the various atmospheric processes that result in precipitation. This study used flow predictions for Ribb and Gumara rivers. Ribb and Gumara frequently inundate the Fogera flood plain, in the Lake Tana area of the Upper Blue Nile, during heavy rainfall events. The study concentrated on assessing the performance of different QPFs that are used as input in the hydrologic model for flow prediction. HEC-HMS hydrologic rainfall runoff model calibrated and validated with data from 2000-2006 was used for flow prediction. Six precipitation forecasts namely: - Zero, Monthly Mean, MM5, MaxEPS, MeanEPS and MinEPS were used as an input in the hydrologic model to predict the flow in Ribb and Gumara Rivers with up to three days lead time. The performance of each precipitation forecast was indirectly checked by comparing the observed and the predicted flow over the two rivers. A total of seven peak flow events in Ribb and Gumara rivers were used for analysis. Each precipitation forecast was analysed based on how often and by how much the peak flow was over/under predicted, how often/how well the precipitation forecast was able to capture the rainfall signal and the relative error in peak flow prediction. It was found that all QPFs, namely: - MM5, MaxEPS, MeanEPS and MinEPS were able to capture the rainfall signal passing over the catchment for most lead times. MaxEPS and MM5 precipitation forecasts in most cases resulted in over predicted peak flow. However it was observed that, for the parameterization settings used, MM5 forecasts resulted in excessive over prediction of peak flow for some events. MeanEPS, MinEPS, Monthly Mean and Zero precipitation forecasts gave under predicted peak flow for most events. The analysis using the relative percent error for peak flow prediction has shown that MaxEPS and MeanEPS forecasts perform better than the others. Finally, for this case study, MaxEPS precipitation forecasts were assessed to perform best for flood warning purpose because they gave flood warning signal for most events.