Rainfall-runoff modeling for South Pare-Usambara Mountains Catchment in Pangani River Basin, Tanzania.

Different Rainfall-Runoff models varying from simple system models, lumped onceptual to Semi-Distributed Physically based were applied for South Pare Usambara mountains catchment, Southeast of Pangani River basin, so as to test the applicability of various hydrological models in the region.One of the applications of Rainfall-Runoff modeling is for extension of discharge data series. Along side with reconstruction of streamflow data in Usambara and South Pare catchments, the system type of models (Total Linear Response and Linear Perturbation form of Linear Transfer function model) were tested for their applicability. As a result, this type of models has shown low model performance, performance index values of 31% and 58% being the lowest and the highest respectively. Lumped conceptual models, namely SMAR, XINANJIANG and NAM were also applied and tested in the catchment under study. There were a significant increase in the model efficiency for some stations (e.g. 1DAlA,R^2 of 76%), although other stations exhibit very low index values. Other performance indicator values like the Mean Square Error (MSE) and the Ratio of Estimated Mean to Observed Mean has proved that the applied conceptual models have the capacity to maintain annual runoff volumes. Finally, a Semi-Distributed Physically Based hydrological model developed for Pangani River basin was used to simulate daily discharge time series. The model has given very low R^2, even negative values. But the simulation of monthly discharges is better than the daily discharge simulation. Analysis of model results and investigation of the failures of those models to simulate daily discharge were carried out. In addition to some reasons associated with the, inherent structure and nature of the models, it was found that the major reasons accounted for the failures and inapplicability of the system type and conceptual models in the watershed under study is due to the persistent anthropogenic disturbances such as irrigation abstraction, deforestation and land degradation that are prevalent in the area. Due to the complexity and heterogeneity of hydrological processes and the simplified assumptions and empiricism used to express the physical phenomena, the results of the Semi-Distributed Physically based model were not appreciable. Such models have not yet reached a convincing stage to be used in operational hydrology in preference to traditional empirical models. Based on the application and analysis of a variety of hydrological models in the study area; It is believed at this stage that unless more representative fully distributed Physically based model is developed for the region, Lumped Conceptual models can be used as better forecasting tools on daily basis for some sub basins in the catchment.