Long Term Rainfall-runoff- Lake Level Modelling of the Lake Naivasha Basin, Kenya

The Soil and Water Assessment Tool (SWAT) model, coupled to a GIS, was applied to the Naivasha basin in Kenya, a closed basin experiencing diverse climatic conditions from semi arid to humid, comprises with the only fresh water lake in the grate rift valley. The intent of this study was to estimates the lake water level fluctuations by integrating the SWAT with a water balance model, estimating the lake water level based on the lake volumes. During the basic data preparation stage of the study, the land use map and the digital elevation model covering the study area was derived with the help of remotely sensed information (TERRA-ASTER). The soil map of the area was updated with the soil physical properties, estimated during the study as well as the information found in the literature. Stream flow analysis was carried out for different areas in the basin to understand the basin hydrological responses, especially the behavior of the base flow. Weather generator, which uses to generate daily rainfall was modified by introducing repetition and adjustment procedures proposed in this study. This modification refined rainfall simulations in terms of the monthly totals. The effect of this new modification is presented in chapter 4. A combination of trial and error and automatic methods to use calibrate the model using the observed monthly stream flow from 1965 – 1975 period. The Parameter estimation program (PEST) was used for the automatic calibration. Manual calibration was performed until the Nash_Sutcliffe coefficient become grater than 0.5. After that the SWAT model was integrated with the PEST program and the Automatic calibration was carried out. Pre sensitivity analysis showed that curve number (CN), groundwater “revap” coefficient and saturated hydraulic conductivity of the soil (SOL_K) are the most sensitive parameters to the model. In addition to that available water content of the soil layer (AWC), ground water recession coefficient ( gw α ) and threshold depth of water in the shallow aquifer for return flow to occur (GWQMN) was assigned to the automatic calibration. Because of the high correlation with revap, AWC was omitted from the optimisation. Other 5 parameters showed low parameter uncertainty of the optimisation and the higher sensitivity to the model. Automatic calibration resulted the further development of the association between observed and simulated stream flows. Validation period was selected as 1935 – 1975 and calibrated model was applied to simulate the stream flow during that period. Results showed the good agreement between observed and simulation stream flows after adjusting the CN number by –7%. Monthly stream flow simulated from 1935 – 1998 was integrated with the EXCEL spread sheet based water balance model to estimate the lake water fluctuations. Root mean square error calculated between estimated and modeled lake water level during 1935 1980 showed 0.72 meters. This indicates the model performed with an acceptable accuracy. Further the modeled lake levels, based on SWAT simulated stream flow, were compared with the lake levels estimated by other two models developed in previous studies and showed better performances.