The effect of hydrologic model development and calibration complexity on prediction accuracy

Abstract

Hydrologic modelling is central to the solution of many water quantity and quality issues. As the complexity of these issues increases, model complexity increases. The purpose of this research was to determine the effects of model complexity on hydrologic model prediction accuracy. Model complexity can enter through the formulation of the model structure as well as the selection of calibration data and criteria. A complex hydrologic model was developed and then simplified based on structural complexity and the change in accuracy was assessed. The results showed that complex models containing excessive low sensitivity parameters did not significantly improve prediction accuracy. However, a lack of complete representation of the physical processes of the hydrologic cycle did affect prediction accuracy. Anomalies in the calibration data can suggest poor prediction accuracy regardless of actual model prediction capabilities. And calibrating to meet specific design criteria should be avoided to ensure optimal overall prediction accuracy. Guidelines were developed to improve future development and application of hydrologic models.