Initial calibration of a semi-distributed rainfall runoff model for the Congo River basin

A monthly time step hydrological model was applied with the aim to address the challenge of water resources estimation in the Congo basin. This paper presents the results of an initial calibration of the model in the basin. In the absence of adequate local input rainfall data for the model, global datasets (CRU TS 2.1 and GHCN-v1) were explored as possible surrogates. Sixty sub-basins were delineated on the basis of the main tributaries and the location of the stream gauges. The initial parameter values of the model for the sub-basins were established using a Monte Carlo sampling approach, which also accounted for model parameter uncertainties. Multiple objective functions were used to determine behavioural and nonbehavioural outputs based on comparison with observed flow data at 16 gauging sites. Manual calibration was then used to establish the regional scheme of model parameters for the basin, including the ungauged catchments. Assessment of the model performance indicates that the hydrological behaviour of the natural system has been fairly reproduced for the major parts of the basin. However, the model application in this study has revealed the importance of exploring the behavioural parameter sets in the context of the available basin physical property data (topography, drainage patterns, geology, soils, vegetation, etc.) in an attempt to constrain the plausible parameter sets to those that are conceptually realistic and consistent with real hydrological processes. The model application has also revealed the discrepancies in behavioural parameters as a result of uncertainty in rainfall input to the model.