| Summary: | Thesis (M.Eng.)--Memorial University of Newfoundland, 1999. Engineering and Applied Science Bibliography: leaves 138-141 Rainfall-runoff models used for hydrological modelling usually involve many parameters that must be calibrated using observed rainfall and runoff data before they can be used for any water resources study. Traditionally, calibrations of these models are done using a trial and error approach or by using numerical optimisation methods, neither of which is entirely satisfactory. In this thesis, a calibration based on Statistical Experimental Designs and Response Surface Methodology is presented. This method integrates statistical experimental designs, regression modelling techniques, and optimisation methods in the calibration process. This method can effectively select the parameters and indicates their interactions that will significantly affect the response variable, which in this case is a goodness-of-fit criterion. The method also determines the optimal values of the parameters that should be used in the model to produce the best fit of calculated runoff amounts to observed runoff amounts. Full factorial and fractional factorial designs and two popular response-surface designs: central composite (CCD) and Box-Behnken were compared. -- Mock's rainfall-runoff model, a popular model for irrigation planning in Indonesia will be used to illustrate the proposed methodology. It has six parameters to be calibrated from observed monthly rainfall and runoff data. The results of the proposed methodology of calibrating the six parameters of the Mock model will be compared to those already obtained previously using the trial and error method. Observed rainfall-runoff and evapotranspiration data from 1973 - 1976 for the Babak River Basin in Lombok, Indonesia will be used in the calibration of the model. Data for 1977 and 1978 will be used for verification of the model. -- The results showed that the proposed methodology gave a better understanding of how the parameters interact with each other, is more systematic, and the optimised values gave a better fit of computed and observed runoffs.
|
|---|