| The successful application of a CRR model depends on how well we handle each phase of model calibration. Despite the popularity of CRR models, reports in the literature indicate that it is typically difficult, if not impossible, to obtain a unique set of optimal parameters for a CRR model. Unless the best set of parameters associated with a given calibration data set can be found, it is impossible to determine how sensitive the parameter estimates (and hence the model forecasts) are to factors such as input and output data error, model error, quantity and quality of data, objective function used, and so on.;In this dissertation, results that clearly establish the nature of the problem of multiple optima in CRR models are presented. Based on these results it is shown why currently used optimization procedures have little chance of successfully finding the optimal parameter sets. This understanding is then used to develop a new global optimization procedure, the Shuffled Complex Evolution (SCE) method, which can efficiently and effectively identify the optimal values for the model parameters.;The efficiency and effectiveness of the SCE method is first demonstrated on some theoretical test functions. It is then used to calibrate a research version of the SMA-NWSRFS model--the SIXPAR model. The SCE method is compared to other available methods used in practice on the theoretical test functions and the SIXPAR model. Finally, the SCE method is applied to the full scale SMA-NWSRFS model using both synthetic data and real data. The test results clearly indicate that the SCE method is superior to other methods tested in this research. |
