Uncertainty Assessment of a Water-Quality Model for Ephemeral Rivers Using GLUE Analysis

Every model is, by definition, a simplification of the system under investigation. Although it would be desirable to reduce the gap between the simulated and the observed behaviors of the system to zero, this reduction is generally impossible owing to the unavoidable uncertainties inherent in any modeling procedure. Uncertainty analyses can provide useful insights into the best model approach to be used for obtaining results with a high level of significance and reliability. The evaluation of parameter uncertainties is necessary for calibration and for estimating the impact of these uncertainties on model performance. In this context, the uncertainty of a river water-quality model developed in previous studies is presented. The main goal is to gain insights into the modeling approaches concerning small rivers. Previous works generally focused on modeling the large river while neglecting the small one. Following a model calibration, the model uncertainty has been assessed by means of the generalized likelihood uncertainty estimation (GLUE). The results showed that the biological process related to the biological oxygen demand (BOD) removal influenced mainly by the parameters characterizing deoxygenation and nitrogen removal. On the other hand, the biological processes related to nitrogen removal were influenced not only by the parameters related to the nitrogen removal but also to oxygen concentration. The application of the GLUE methodology shows that the river quality model considered is suitable for simulating the important processes involved. Uncertainty bounds showed different amplitudes with respect to the pollutant species considered. In particular, the oxygen uncertainty bounds were narrower with respect to the other model components suggesting much attention must be paid to both model algorithms and quality data to be gathered. The study confirmed the suitability of the GLUE methodology as a powerful tool as a simplified screening methodology to assess uncertainty.