Abstract: A new concept of ‘reliability’ is introduced to evaluate the performance of dynamic models. To propose a new method for model reliability assessment, the basic methodology used in the study involves different error estimation methods used in statistical analysis. Performance of the methods is evaluated by applying them to 12 different types of synthetic hydrographs. Before proposing a new method, the performance of 7 existing, widely used statistical error estimation methods is evaluated. As the existing methods failed to produce satisfactory results, 3 of these 7 methods are modified. Even with the modifications, the results of the 3 modified methods were not satisfactory. As both the existing and the modification of these existing methods failed to produce satisfactory results, in search of a single acceptable method, we propose 6 new methods that are based on normalization of the mean of the absolute difference of the instantaneous measured values from either the minimum or the maximum or the average of the measured values. We found that the method based on the average of the measured values produced the best result. We named this new method <i>Normalized Squared Mean Average</i> and abbreviated it as <i>NSMAVG</i>. This method is recommended for use in calculating the reliability of dynamic models. The proposed new method can be used to evaluate the model performance during the calibration and validation of a dynamic model.

Keywords: Dynamic model, Reliability, Error Estimation, Calibration, Validation

Pages: 114-135

WSEAS Transactions on