Search

The goal of this study was to evaluate regionalisation methods that could be used for modelling catchment inflows into Lake Victoria. WASMOD, a conceptual water balance model, was applied to nine gauged sub-basins in Lake Victoria basin in order to test the transferability of model parameters between the basins using three regionalisation approaches. Model calibration was carried out within the GLUE (generalised likelihood uncertainty estimation) framework for uncertainty assessment. The analysis was carried out for the period 1967-2000. Parameter transferability was assessed by comparing the likelihood values of regionalised simulations with the values under calibration for each basin. WASMOD performed well for all study sub-basins with Nash-Sutcliffe values ranging between 0.70 and 0.82. Transferability results were mixed. For the proxy-basin method, the best performing parameter donor basin was Mara with four proxy basins giving acceptable results. Sio, Sondu, Gucha and Duma also performed well. The global mean method gave acceptable performance for seven of the nine study basins. The ensemble regionalisation method provides the possibility to consider parameter uncertainty in the regionalisation. Ensemble regionalisation method performed best with an average departure of 40% from the observed mean annual flows compared to 48 and 60% for proxy-basin and global mean methods, respectively. [ABSTRACT FROM AUTHOR]

To predict rice-farming pesticide concentrations in river water with imprecise model inputs for screening-level analysis, a basin-scale runoff model was developed. The Monte Carlo method was applied to create estimates of input data regarding agricultural work schedules and parameters for pesticide decomposition and sorption in solids and water. The prediction accuracy of the model was evaluated when used with non-optimised pesticide parameters; the model was calibrated using hydrological data alone without reference to observed pesticide concentration data. Overall, predictions for the pesticide concentrations were successful within order-of-magnitude accuracy. The pesticide rankings according to the predicted concentration roughly agreed with those observed. The success of screening-level analysis indicates that the model prediction can help in selection of pesticides to be monitored and in determining the monitoring schedule for the river basin. [ABSTRACT FROM AUTHOR]