Model performance of a conceptual hydrological model over the Nile Basin: effect of parameter uncertainty on the performance of PCR-GLOBWB 2

Abstract

The Nile River is the major river basin in the northeast of Africa with an ever-growing population dependent on it for its water needs. These needs are only expected to grow in future. Predicting Nile discharge is hard as discharge of the river itself can vary a lot from year to year. Furthermore, Global Hydrological Models, such as PCR-GLOBWB2, also struggle to accurately predict Nile flows and tend to overestimate discharge, most likely due to poor data availability. The Nile River itself can be divided into several basins, with the Blue Nile being the most important in terms of discharge, with 60-70 % of discharge originating here, but this comes mostly in the summer months. The rest of the year the White Nile, a much larger basin with its origins in the region around Lake Victoria, is the main source of water. To improve model performance of the PCR-GLOBWB 2 it was chosen to improve input data and parameterization. By using updated data soil data (SoilGrids), topographic data (MERIT-DEM) and meteorological data (WFDE5) and using an updated river routing method. The results were validated by GRDC data, but as these are limited, both spatially and temporally it was chosen to seek an additional way of validating data. In this case GLEAM was used to validate evaporation of the model. Combining of both validation methods was used to better understand the discrepancies in model performance for the Nile Basin. Results still showed poor results for the run without any improvement as expected. It showed that the White Nile performance was very poor, owing to an underestimation of evaporation, mainly in the months of no precipitation. This was observed in some other basin as well. The Blue Nile performance was better but still has overestimations of discharge, especially in the drier months. Validation of the results with GLEAM showed good options to close the water balance and get better temporal and spatial data despite some of its current limitations. Introducing new datasets and new parameterization did not improve model results as expected and the poor datasets were not the main cause of the poor performance. The same regions which performed poorly for the standard run also performed poorly for the other runs, indicating an underlying problem. This has to do with the evaporation in the model, especially in the Sudd swamps, where flooding does not occur as expected leading to larger discharge and decreased water storage. Here it was identified that missing model descriptions is a large problem and that this process needs to be better described. For future, many improvements can be made to model and by this, combined with better validation methods, the description of PCR-GLOBWB 2 of the hydrology of the Nile River could be improved.